Sunday, 15 April 2012

Shared neural resources for throwing and language: a whacky idea for an experiment

For some time now, there has been an hypothesis floating around in evolutionary biology that the human capacity for language emerged, in part, from the development of our ability to throw long distances with high speed and accuracy. There are a few reasons to think this, mostly correlational, inferential kinds of reasons, but they are accumulating.

We were chatting one day about how to test this hypothesis a bit more directly, and we came up with a whacky experiment. We'd like advice from neuroscientists with experience in brain stimulation techniques about whether this sort of thing is feasible. We'd also like to brainstorm the logic of this experiment and see if we can come up with a practical design that stands a chance of finding something. We then need collaborators; I can handle the throwing side (analysis, measurement, etc) but we don't know anything about TMS and would need an expert on board.

There are many other reasons why this might fail, though - I still need to do a detailed lit review on the throwing/language references I have. Our main problem is that we don't know the kind of obvious difficulties in doing TMS in this kind of context. We'd like to assemble a) an experiment and b) a research team to do the experiment if we can get it to make sense, and if it works we will submit the hell out of this to Nature :)

Throwing for long distances with accuracy requires the ability to produce very precisely timed and coordinated action, spread across multiple limb segments.
Figure 1. The basics of a throwing action
High release speed requires a lot of force; the muscles in the hand and wrist are too small to generate the required forces, and so instead we use the large, slow muscles in the torso to develop force during the wind up. This force must then be transmitted along the link-segment system of the arm in a kinetic chain; done correctly, the force remains high and it therefore accelerates the lighter and lighter segments to faster and faster speeds. This force is finally transmitted to the projectile, which goes on it's merry way.

This only works if the timing of the segments taking up the force is controlled precisely. If the segment begins working too soon or too late, some of the force is lost. In addition, if you need accuracy as well, then the final release of the projectile has to occur at just the right time so that it heads in the right direction. There is some evidence that throwing was Homo sapiens' major advantage over competing species. Our brains are smaller than Neanderthals, for example, but we have larger cerebellums and posterior parietal regions, all implicated in timing.

Throwing & Language
The hypothesised relationship between throwing and language is based in this issue of coordination. If we came to develop throwing and this was selected for because of it's advantages in hunting, we will also have developed the neural resources required to support this complex perception-action skill (brains may not be everything, but they are critical resources for any embodied organism). These resources were then available to support other skilled action that requires complex and precise coordination across multiple body parts, specifically spoken language. Speech is a remarkable feat of skilled movement, and the hypothesis is that evolving the ability to throw made the capacity to speak an option, which Homo sapiens then took.

Common neural underpinnings
I recently reviewed a paper that examined throwing and communication in chimps. They performed structural scans on chimps who could throw and those who couldn't, and found chimps who throw have a higher proportion of white matter in inferior frontal gyrus (the chimp homologue of Broca's area) and areas of precentral gyrus compared to chimps who do not throw. This effect was lateralised to the left hemisphere for right handed chimps, while lefties were less lateralised. Finally, the two groups of chimps (throwers vs non-throwers) were only different in one cognitive measure; throwers out-scored non-throwers on a measure of communication skill from the modified Primate Cognition Test Battery. This is completely correlational and the relationship between the communication test and the timing stuff is weak, but it's suggestive and it's a start.

The whacky hypothesis
We started thinking about how to test for this connection in people. It's hard because language skill and throwing skill are mediated by some many varied factors in modern life, there's no particular reason to expect them to co-exist behaviourally. However, the hypothesis is about common neural machinery, so we thought of the following idea.

What if you could knock out, say, Broca's area or something else that's typically thought of as supporting speech. Then, what if, when you tested people's throwing, they were impaired in a way you wouldn't expect unless the two systems overlapped in important ways?

The basic design would be this:

1) Use transcranial magnetic stimulation to temporarily knock out Broca's area or something else suitable.
2) Immediately have the person throw multiple times to a target from 5m, 10m or 15m.
3) Profit!

Movement data analysis: I study throwing and we typically film the throw from the side using a high speed camera. The person has reflective markers attached to each joint (wrist, elbow, shoulder, hip, knee and ankle) and we then digitise the video to extract the kinematics of the markers. We use these position time series as sensitive measures of throwing performance (we can measure how the joint angles relate to one another and the timing along the segments, for example). We can also compute release angle and velocity, the two key parameters the thrower is trying to control; in experts, these vary systematically with target distance and height as the expert tunes their throw to current task demands.

In short, if we had expert throwers (who can produce stable throws) we could detect subtle changes in the throw kinematics after TMS, if there are any.

Here's where we need help
There are numerous questions about TMS and whether this design makes sense we need advice on from anyone with the relevant expertise,so please comment away! We have no particular ego invested in this basic design, it's just our starting point for a whacky test of something kind of cool, so feel free to go to town.

1. TMS: According to Twitter, hitting Broca's area is possible but can hurt, as it triggers facial twitches. What are the ethical and practical issues around getting people to agree to this -is it hard? This is particularly complicated as we need decent throwers - novice performance is already too variable.

2. Apparently continuous theta burst stimulation should knock out function for 10-15 mins. Is this complicated? How well can you target it? What do you need to target it?

3. We're predicting that knocking out a language related part of the brain will have detectable effects on throwing because these tasks rest on shared neural resources for coordination of precisely timed actions. We're starting with Broca's area because of the chimp study. Are there other areas worth thinking about?

4. Would anyone be surprised if we got motor effects with this kind of stimulation? I'm thinking we might need a dissociation task, where we do TMS, find it affects throwing but not some other fine motor task. Thoughts?

Thoughts and comments welcome!


  1. Sounds pretty cool. Before going with the TMS, it might be worth trying some sort of behavioural interference study (i.e., does a language task interfere with throwing more than a right-lateralised task like number or face judgements?). Plenty of reasons why that might not work, but if it did, then your TMS story would be a lock.

    1. I doubt recruitment will be an issue - people who are interested will take part regardless of how painful you tell them it will be, but a good chunk of your (non-undergrad) sample will not particularly like the idea of having part of their brain turned off. I suspect you'll have more of an issue with drop outs than with recruitment. But perhaps the competitive nature of your sporty sample will deal with that problem...

    2. Our current ethics protocol doesn't allow theta burst (I suspect this is just because no one has had any reason to do it yet, rather than any ethical issues), so I've never used it myself. For all TMS studies, you need some way of localising the brain area. The best (and consequently most expensive) is to localiase the area with an fMRI task (word imagary, presumably, would light up broca's), and then use a co-registration system like Brainsight to make sure you are targeting the coil at the right area all the time. This is the gold standard, and if you have the money and facilities, this is the way to go, but certainly not necessary for publication. More likely will be to find out the coordinates that generally used when broca's is wiped out with TMS, and resign yourself to a lengthy hunt with a simple work knock out/stimulation mini-experiment, or just pray for minimal variability between subjects.

    3.I guess it might be nice to show that knocking out a motor area (SMA for complex sequences of movements? Some sort of premotor region?) affected throwing in a similar (or, better yet, different) way, kinematically? SMA is also challenging to hit since it is on the medial wall and requires a fancy coil.

    4. Even more than a control tasks, I woudl say you need to provide sham stimulation to a location that will be as distracting as the target areas (i.e. just as painful), showing that it doesn't affect throwing (to rule out trivial effects of TMS like the blinking and noise and weirdness on a task). But maybe this is less of and issue with theta burst... I guess for a control task you could just examine some non-kinematic chain actions like peak grip strength or rate of finger tapping?

  2. Hi guys, interesting idea. The TMS bit is pretty straightforward; the devil is more in the behavioural task and choosing appropriate controls.

    This is how I would do it.

    1) localize Broca's area using a TMS speech-arrest task. This has two advantages: first it is cheaper than fMRI and, second, it guarantees that your TMS is targeting a functionally critical region. This is easily done by identifying left pars opercularis in each subject's structural MRI scan, then mapping a nine-point 1-inch grid around that area. You ask the subject to count as quickly as possible from 1-10 in a repeating cycle while giving 5Hz stimulation at about 140% distance-adjusted motor threshold (or less depending on comfort threshold, obtained in an initial session). Then define your stimulation site as the region that produces the most pronounced speech arrest (through speech errors) or the most impaired counting time (through number of cycles completed), calculated relative to a sham TMS condition.

    2) Now that you've defined Broca's area you need a control site. Best would be nearby but beyond the area that induced speech arrest; that way you'll get similar twitch artefacts and discomfort. Left IFJ could be good, although it has been associated with task-switching (see work of Marcel Brass; need to choose the control area carefully based on the literature). Corresponding area in the right hemisphere also a possibility (e.g. right pars opercularis) although this has been linked to response inhibition and attention, so could influence throwing ability. Also include a sham cTBS condition as a behavioural baseline. Include site as a within-subjects factor in the design.

    3) Need a control task that is as similar as possible to throwing but doesn't tap into the precise motor/cognitive mechanisms that you want to test. Need to match carefully in terms of extent to which the task draws on attention, exertion, arousal etc. and controls for possible effects of TMS on these processes. May need multiple control tasks to catch different processes. e.g. you probably need to confirm that cTBS influenced something to do with speech, as it is conceivable that it might not (due to lower intensity stimulation of cTBS compared to the hunting protocol in point 1)

    4) In general the effects of continuous TBS last about 45+ min, so plenty of time to do multiple tasks in one session. Need careful counterbalancing of sites and tasks across subjects. The effect of cTBS can take 5min to peak (see Huang et al 2005, Neuron) and on top of that, we've found the timecourse to vary considerably between subjects (Verbruggen et al 2010, PNAS) so important to have multiple samples within subjects over a period of 60+ minutes. Can include time after stimulation as a factor in the design, or do linear mixed effects analysis to model individual differences in the timecourse.

    5) Ethics and recruitment: should be straightforward. Many labs now have approval for the patterned protocols mentioned in Rossi et al 2009 (Clin Neurophys). TBS is given at 80% motor threshold so theoretically has a lower seizure risk than high frequency rTMS protocols given at a higher intensity. There is only one documented case of seizure following cTBS (Oberman et al), and it was applied at a higher intensity than standard.

    Having said all that, I'd start with a basic pilot in 6 subjects of sham cTBS vs Broca's with just the throwing task. If you don't show an effect in at least 4/6 then rethink. No point running a big complex study full of control tasks etc. unless you have a sense that the TMS has an effect to begin with on the main task of interest.

  3. Gavin said,
    "it might be worth trying some sort of behavioural interference study (i.e., does a language task interfere with throwing more than a right-lateralised task like number or face judgements?)"

    Great idea, and it should be pretty easy once the conditions are nailed down.

    Obviously we need two dependent variables, a throwing task and some other motor task - how about controlled dropping (on a target)? Controlled dropping would have the virtue of using pretty much all the same bodily systems, but being very robust to the micro-coordination required for throwing.

    It would be nice to have three conditions in which participants do these tasks - 1) language generation, 2) language comprehension, 3) potentially-distracting non-language.

    How about:
    1) 'Tell story about recent events' or 'tell me everything you remember since arriving here'. (Many other language tasks might seem good, but we need something that produces continuous language generation, so alternating games like "rhyme this word" probably won't work.)

    2) Listen to story you will be expected to answer questions about, either on recognition or recall task. (This control will indicate whether it is specifically language generation that interferes with throwing vs. any language task.)

    3) This is harder, a non-languge task you can do while throwing. How about a haptic memory task done with the off hand. Some sort of pattern of finger movements is presented visually or haptically, and the pattern must be repeated back?

    Now, there is one problem in this design, which is not in the TMS experiment: If there is interference, it could be either direction. That is, in condition 1, we might find that skilled throwers remain highly accurate, but they stink at telling stories while they throw. That would be a great finding, but it would have to be measured somehow.

    What do you guys think?

  4. I think technically it could be done with cTBS followed by a throwing task or other task. But if you showed the ideal result, ie. cTBS of IFG ==> worse throwing and worse language, I'd still worry that the effect was not specific to throwing. The challenge would really be to show the specificity of your effect. Does TMS just make you worse at everything?

    So you need some really good comparison tasks, matched for motor difficulty etc but without the critical components that you think matter in throwing. And of course, the more tasks you have, the harder it is to do your TMS (you start needing multiple sessions over multiple days). So I think the behavioural interference task would be a much better place to start.

    Another way to look at this question would be to look developmentally. Do kids with poor motor skills also have poor language skills? I'm sure there is data on that somewhere.


  5. My suggestion is to include people with aphasia as participants in your research. You can identify participants with damage to particular areas of the brain (e.g., Brocas). You can assess their language abilities using standardised language tests. What impact will throwing have on language rehabilitation for people with aphasia? You could compare control participants (with no damage to language in the brain) to participants with aphasia, who are diagnosed with language difficulties due a stroke etc. Also, how about analysing and comparing throwing arm usage with control participants, participants with aphasia and hemiparesis, participants with aphasia and hemiplegia, participants with aphasia and no weakness/paralysis.

  6. [sorry this is a bit behind the curve - couldn't post it yesterday!]

    Firstly I want to respond to the practical issues about using TMS in targeting Broca's area because this is not a particularly difficult thing to do but it is one that has some practical problems.

    One problem will be that the TMS coil needs to be placed very carefully because TMS is a highly focal technique. If it's not placed on exactly the right point on the scalp you might be hitting the wrong piece of brain. Usually this is accomplished by having the subject sit in a frame, with the coil fixed to the head.

    This would make an online TMS experiment impractical. The head movements from the throw would mean that the coil would have to be relocated on each trial.

    The alternative would be an offline stimulation method, such as theta burst stimulation (TBS). In this case you would measure kinematics before the stimulation, then do the zapping, then measure again. TBS involves delivering a large number of pulses quite rapidly (within a minute, usually). The temporal pattern of pulses may have either inhibitory or excitatory effects, depending on a number of factors (Huang et al., 2005). You would need some kind of sham stimulation condition to rule out practice effects.

    I would be tempted to look into transcranial current stimulation tCS comes in two flavours tDCS which is usually an off-line technique like TBS or tACS which may also be used online and is used to deliver specific frequencies of stimulation. tCS has the advantage that the stimulating box is relatively small and easily portable, and the electrodes that deliver the current are also small and can be fixed to the head with rubber straps. The stimulation is less focal than TMS, but is much more subject-friendly!

    Huang et al. 2005:

  7. the connection is that both throwing and sound production are complex motoricaly. What about comprehension? Why throwing and not some other activity? The idea of a evolutionary connection seems quarter-baked at best.

    1. These are good questions. There are, as I understand it, a few converging (but only correlational) lines of evidence

      1. Throwing and language were only really developed fully in H. sapiens

      2. They show up around the same time and are related to H. sapiens-specific changes in cortex

      3. The action aspect of each task involves this precise coordination and timing across segments in ways that most other tasks don't; skilled action is good but often operates within slightly more relaxed timing constraints.

      4. This chimp study is suggestive of a quite direct overlap.

      There's more that I haven't had a chance to read in much detail; Bingham & Zhu cite a bunch of it (Paper here).

      Yes, it's very preliminary so far - hence our whacky idea to just test the thing directly :)

    2. This is how evolutionary hypotheses are supposed to work. There is a convergence of evidence, slow at first. A hypothesis is made, and then people go about testing it.

      Certainly this hypothesis is not supported well enough that the whole field of psychology should flock to it, but it is supported enough by evidence from evolutionary biologists and biological anthropologists to warrant some direct tests. The most aggressive prediction you could make from the hypothesis is about brain areas, but there might be a range of behavioral and developmental predictions you could test as well. Again, this is exactly how it should work.

      The only non-ideal bit, from an 'evolutionary theory' point of view, is that there does not seem to be a proposed connection between throwing/language and niche matching. This will not bother most people though, and is totally separate from the brain-area-usage question.

    3. The idea of an evolutionary link is quite interesting, considering that language as we know it today is but an extension of communication, motor or vocal. Before vocal language, I would bet that a physical and more motor form of communication dominated tribes of man. I would absolutely say that the areas are linked and that the effect would not be limited to throwing but to any 'dramatic' motor movement. It would be interesting to see whether language and dancing have similar neural networks...because if I'm right in my thinking and that motor movements dominated communication before vocal language and that it is the same neural networks just being expressed differently, then dance which is also a form of communication may have a similar network also. Interesting ideas!

  8. Thanks for comments all! Some thoughts;

    My instinct would also normally be to run a behavioural task rather than jump right in at the neural deep end. My gut feeling, though is that we'd never see speech specific interference on throwing; the overt behaviours are supported by too many things (practice, culture, etc etc). The reason to go to TMS is cut to the chase; the hypothesis is about shared neural resources, let's go there. Not to say I couldn't try a dual task, but I wouldn't put any money on it working.

    Antonia: a developmental angle is not a bad idea. That said, behavioural measures will all be messy; Kids with DCD, for example, would be bad at the throwing but they'd be bad for lots of reasons that aren't specific to throwing. You'd have to track the development of throwing and language skills and see if they tracked at all - are good language learners good throwers? Interesting, but long term.

    Aphasics with known lesion sites around Broca's is interesting - I could try training them, see what happens. The goal of this project is to effectively do a lesion study with participants I have access to, though so this isn't more practical. Still, if anyone has some Broca's aphasics handy, let us know! :)

    Nick: I asked this on Twitter, I think; as I understand it, tDCS is about altering plasticity. So people tend to talk about it in terms of changing learning rates, etc. Can you use it to do the kind of faux lesion that TMS/TBS achieves, becuase I still think this is what's needed; knock out something speech related and show surprising effects on throwing.

  9. Another (side, almost tangential) issue that you might want to consider in the method is refraining from using verbal instructions since chances are language comprehension is affected in this paradigm.

    Also, given your rationale on shared neural circuitry between language and throwing, would you predict impairment in language function (specifically, syntactic information processing) when similar TBS/TMS is performed on the motor cortex (or other brain region controlling throwing motion)?


    1. Hm. Good point on the instructions, although as I understand it we're aiming for speech production regions, rather than speech comprehension (yes I know I'm glossing over a lot of how the brain works there :)

      About the reverse results - yes, presumably, but I'm not sure it's quite symmetrical. There are regions of cortex known to be basically about language and speech production, but as far as I know, nothing in, say, motor cortex specifically about throwing. Cerebellum is the obvious target, because of it's role in timing, but again it's not that specific.

    2. My concentration is in psycholinguistics, so I am naturally sensitive to these measures. I am picturing reviewer going "so you didn't control for this - denied" if the participants are "primed" with language in any form - production, comprehension or whatnot. Then again, yours truly is just...sensitive.

      At the same time, my current research direction (as a graduate FYP) is on graspability and graphic production, so I am inclined to say that we are in the similar boat as I am looking for linguistic influence and handedness effect using biomechanic sensitive task.

      Let me just say that it is hard to control for so many factors.

      Anyway, I do want to add that we know that gesture is integral in communicating to the point where it is used even when people are on the phone looking like an idiot waving hands and all (cite: personal experience). I am confident that somewhere in the literature, studies have been conducted on restraining participant's movement vs. not restraining and test for difference in linguistic task. So look it up and see if you can find something?

      I'll be scouring the literature this or next week on neuropsychology of bilingual mind (or bilingual brain, whatever :) )for my final paper. I'll update you if I find anything interesting.

  10. I assigned a student in my lab to brainstorm behavioral studies, which were discussed at our Friday lab meeting. We thought an A-B vs. B-A study would be better than a behavioral interference story (throw->speak, or speak->throw). Think of it as a priming study, with the first condition also acting at the control condition.

    By next Friday my student has been told to identify several potential measures: simple throwing measures including cost estimates, and speech generation tasks. We might even try running something over the summer. This is exactly the type of simple study our undergrads are good at executing.

    What do you think? I know for your study you were thinking about highly sophisticated measures of throwing skill, but we don't have the resources for anything like that. I'm also worried about speech generation tasks, and am naively hoping there are established tasks in the literature.

    1. I'd want to measure kinematics because I'm assuming we won't find massive effects; kinematics from expert throwers are pretty stable and small variations show up nicely.

      I'm always interested in more ideas, things we might actually be able to run to show proof of concept. As it is, this kind of groovy study idea is utterly unfundable, because funding is bullshit, so if we were able to get any toe hold into this, that would help.

  11. Hey, a paper is out on PubMed today that looks like it overlaps with what you're planning! Esopenko et al from Toronto - A neuroanatomical examination of embodied cognition: semantic generation to action-related stimuli.

    1. Thanks! I need to have a look at that (I saw you link on Twitter, just haven't had time to reply). It's not our flavour of embodiment, too heavy on the Barselou, but the implementation could be useful.

  12. I really don't have any relevant expertise in doing this kind of experiment, but I study linguistics so I'm very interested in the hypothesis and what you might learn :)

  13. Great to suddenly come across this blog. I performed an experiment on screaming and hitting! Showing that impact velocity of the hit is influenced by words like "near" and "far". I performed the experiment many years ago, but I haven't published it yet. (It's under review)