Understanding Motor Control: Experimental Design Insights
School
University of California, Davis**We aren't endorsed by this school
Course
ANT 20
Subject
Mechanical Engineering
Date
Dec 10, 2024
Pages
10
Uploaded by DeanSnakeMaster1939
NPB 100L Motor Control and Perception LaboratoryObjective: To improve skills in experimental design and hypothesis testing through analysis of a motor control task. To learn how to quantitatively characterize behavior in a classic perceptual task design involving a choice between two alternatives. I. Background A. Motor control and efference copy of motor commands Each time we move a limb or a muscle, signals about this movement are relayed throughout the nervous system. For example, consider what happens when we move our eyes to the left. If our visual system were unaware of this movement, we might think that objects originally projecting on the centers of our retinas had now shifted to the right (since, after moving our eyes leftward, they would now project upon the right portions of our eye). Thankfully, our sensory systems can resolve this ambiguity as to whether our eyes moved or the world moved by receiving feedback from the eye movement system informing us that our eyes have moved. Likewise, when we perform complex motor tasks involving the coordination of multiple body parts, our brains and motor systems receive constant feedback about the locations of our various body parts. For example, suppose we would like to reach for an object that is straight in front of us. If we twist our body while reaching, then our arm movement will need to be adjusted in order to compensate for the twist of the body. In order to perform this compensation, the nervous system has a few strategies: 1. Proprioceptive or other sensory feedback. Proprioceptors such as muscle spindles and golgi tendon organs provide our nervous system with sensory information about the positions, velocities, and loads (forces) being applied to our muscles. These signals can be conveyed very accurately. However, they only arrive following a propagation delay. Thus, for coordinating movements that are happening together or in rapid succession, this information is often too slow to arrive. Likewise, other forms of sensory feedback about our movements –such as those provided by our visual system observing our movements –are also very slow (typically even slower than proprioception). 2. Efference copy. To obtain an estimate of where our limbs or other body parts will be beforefeedback from our sensory systems have time to tell us exactly what has happened to our limbs, our nervous system requires signals that predicthow recent motor commands will affect the future locations of our limbs. To do this, a copy of the motor command is sent to other parts of the nervous system that need to be coordinated with the commanded movement. Such a copy is known as an “efference copy” (recall from our reflex laboratory that “efferent” refers to signals going out to the periphery). With this copy of the motor command, our nervous system can predict the sensory information that would be provided by the proprioceptors (sometimes termed the “sensory consequences” of the motor command) even before our sensory receptors have time to send it! For this prediction to be accurate, our nervous system must contain circuitry that reliably converts the received copy of the motor command into an estimate of the sensory consequences of the motor command. In most cases, we do not know exactly what circuitry performs this
2 conversion from efference copy to predicted sensory consequences. Motor control researchers therefore denote the circuitry that converts efference copies of motor commands into predicted sensory consequences by a “black box” (indicating that the contents aren’t known…or at least aren’t going to be discussed) known as a forward model1: Forward models are not necessarily correct. For example, consider what happens if your motor cortex sends out a command to move your foot forward, but then an object blocks your foot from actually moving. In this case, the forward model would incorrectly predict that the foot would move forward and convey this incorrect prediction to your postural systems. As a result, you would trip because your body isn’t properly adjusted for this unexpectedconsequence of your motor action. Thankfully, our proprioceptors usually are fast enough to trigger tripping-related reflexes that keep us from falling (but not always…). Another interesting feature of this example is that, when the sensory consequences of our motor command from the forward model do notmatch the (delayed) actual sensory signals from our proprioceptors, the mismatch between the predicted and actual sensory consequences or violation of expectations acts as an error signalthat triggers brain plasticity mechanisms and leads us to learna revised set of actions and/or revised forward model. This type of example is at the heart of the field of motor learning.B. Violation of expectations vs. Bayesian combination of evidence and prior experience The above discussion of motor learning brings up several important issues related not only to motor systems, but also to our sensory perceptions. First, in many sensory and perceptual systems, neurons most strongly convey violations of expectations. The simplest example of this is that our bodies commonly ignore sensory signals that do not change in time, but respond strongly to changes –one can think of these changes as “violations” of the most common thing that happens around us: nothing changing at all (e.g. walls don’t moveor change color, etc.). More generally, we have expectations about most events in our day, e.g. our lab will be held in its usual room and will be on the topic of the previous lecture. A second, sometimes competing feature of our sensory systems is that they can combine multiple sources of information in a probabilistic manner and also use prior experienceto make judgments. For example, a trick many grade school children learn is to mouth the words “olive juice” to their classmates. Since the lip movements when saying this are quite close to those of “I love you”, the nervous system needs to combine ambiguous information (lip movements, which aren’t very reliable for telling what word is said)with prior experience (relative likelihoods of someone saying “olive juice”vs. “I love you”). How the nervous system makes 1One also can ask the question: “If I want to achieve a desired set of sensory responses, what motor command should I generate?” Motor control researchers use the term “inverse model” for the circuitry that converts a desiredsensory consequence into an appropriate motor command. Inverse here refers to the fact that the arrows in a black box diagram would go in the opposite direction from those shown above, from (desired) sensory response to (required) motor command. Both forward and inverse models are sometimes referred to more generally as “internal models” since they refer to computations done internally within our nervous systems.
3 probabilistic judgments is a hot topic in neuroscience, known as either Probabilistic or Bayesian decision-making. The key underlying concept from probability theory is Bayes rule: Prob(|)Prob()Prob(|)Prob(|)Prob()Prob()EXXXEEXXE=where Prob denotes “probability”and | denotes “given that”. In our example above, X is the actual word that was said (e.g. we want to know what is the probability that X = “I love you”) and E (the “evidence”) is what our senses received. Thus, to calculate the probability that the speaker said “I love you”, given what we saw, the formula above says: Prob("I love you" was said | Lip movements we saw)Prob(Lip movements we saw |"I love you" was said) Prob("I love you" is [ever] spoken)Note here that the left side of the equation is what we want to know, i.e. did the speaker say “I love you”given what we saw. This is proportional to two factors: first, the probability that the lip movements we saw occur when the words “I love you” are said –this is known as the likelihood of “I love you”. The second factor is how probable it is (independent of what we are currently seeing) that the words “I love you” are (ever) spoken –this is known as the prior probabilitysince it corresponds to our overall prior experience with how often an event occurs (I love you is spoken) regardless of the present situation. In our case, we next would want to compute the corresponding probability for whether the speaker said “olive juice” given the lip movements we saw and thereby determine which phrase is more likely to have been said. Note that the reason this trick works is because “I love you” is a much more common phrase than “olive juice”, i.e. has a higher prior probability (2ndterm on the right-side of the equation), even though the lip movements we saw are more probable when “Olive juice is spoken” (first term of right-side). So…what’s the punchline here?The punchline is that different brain systems may do either of 2 somewhat contradictory-seeming things: If the nervous system is conveying the probability of events and using Bayes rule, then very common events(i.e. those with a high prior probability) will be represented most strongly. On the other hand, if the nervous system is conveying violations of expectations, then unlikely eventswill be conveyed most strongly. Of course, the nervous system comprises many parts and different parts of the nervous system may convey different features of received information. We will explore some of these issues in this lab.
4 II. Equipment setup and testing Make sure that your setup has the following equipment:String potentiometer(on blue board)9 volt battery Velcro strap to attach to handWeighted box Large petri dish Eight-channel analog-to-digital converter and USB cable One BNC –Red and black grabber cable Laptop for recording dataThe string potentiometer should have the battery installed and be plugged by BNC connector into Ch 0 of the A-to-D Converter box, which should be plugged into a USB port of your laptop. Run TracerDAQ Pro, open the oscilloscope, and load the SingleChannel.osc Named Configuration. Set the VOLTAGE SCALE to 1 V, the TIME SCALE to 500 ms, and the TRIGGER mode to Auto. Slide the Ch0 baseline marker down so that your baseline trace is most (but not all) of the way to the bottom of the oscilloscope screen (we will primarily be recording positive voltage changes from this baseline). The string potentiometer records the position of the cable wire that is attached to the strap. Move the velcro strap up and downand see if you can observe a deflection.III. Experiments A. Motor experiment 1: Measuring deflection to self vs. externally generated inputs The goal of this experiment is for you to test hypothesesconcerning the control of motor output and coordination, and to design experimentsto test these hypotheses. 1. Stabilizing the hand during self-removal of object. The subject should place his or her dominant hand in the Velcro strap, with the strap located just below where the fingers attach into the palm of the hand (Figure 1). The arm should be held out comfortably, with the palm upright and hand opened to make a flat surface. The petri dish should be placed on the center of the open palm (Figure 2). The arm should remain as relaxed as possible throughout today’s lab.
5 Figure 1: Arm posture and placement of velcro strap Figure 2: Box placement on petri dish tray One person (the “data recorder”) should be located at the computer and have pencil and paper ready to record the data, which should be out of view of the subject. The other person (the “tester”) will place the box on the petri dish. If there are only two members in your lab group, then the data recorder will need to also act as the tester. Upon the data recorder’s signal, the tester should place the box near the center ofthe petri dish. The subject should wait 1-2 seconds for his/her hand to become stable and relaxed, and then quickly and smoothlylift the box straight up and off of the hand while trying to keep the hand as still as possible. The subject should then give the box back to the tester, relax (letting the arm down so as not to get too tired, given that we will do this many many times), and wait for the data recorder to complete the measurement of the hand’s deflection.The data recorder should measure the initial voltage deflection on the oscilloscope by using the two vertical markers (note: this deflection should only be ~100-200 ms long; do not measure the subject’s later movements after the initial deflection). If there is a single deflection (either positive or negative—please note which!), measure the voltage difference from just before the movement to the peak of the deflection just after the movement. If there is a (small) negative dip before a positive deflection, then measure the size of the jump from the bottom of
6 this dip to the peak of the positive deflection (but do not get this positive deflection confused with the subject making later movements after the initial deflection is already over). Repeat this about 5 times or until the subject is producing reasonably smooth and consistent removals. Then record the individual and average deflection(s) for an additional 5 trials (if there are any very weird/irregular deflection trials, feel free to add another trial). 2. Stabilizing the hand during removal by another person. The tester should place the box on the petri dish, and then instruct the subject to close the subject’seyes. The tester should quickly but smoothly lift the box straight up to remove the box from the petri dish and record the deflection. Perform 3 practice trials, and then record the following 5 trials. 3. Repeat steps 1 and 2 for a second subject/lab group member. 4. Hypothesis development.Was the deflection larger when the subject removed the object or when another person removed the object? Develop a set of hypotheses for why the deflection was different in the two cases above. Record your initial hypotheses on the following (blank) page. 5. Experimental design.Design tests of your hypotheses. Systematically note what you do to test these hypotheses, and what the effects of your various experimental manipulations are. Continue to refine your initial hypotheses as necessary. 6. Challenge(!): As you continue to refine your hypotheses and put your ideas into action, how close can you get the size of the deflections when another person removes the object to approach the size of the deflections when the subject removes the object him/her/themself? Come up with the best strategy you can and report it to an instructor. You can choose a single person to be the subject for this part, but all lab members should contribute to developing hypotheses and how to test them.
7 [Blank space for hypotheses and strategies for testing these]
8 C. Threshold psychophysics In this part you will perceptually compare the weights of objects that are either the same or different sizes. For this, you should have: • 1 medium, 1 small, and 1 large “Standard” boxes; the medium-sized Standard box is labeled with 3 colored stickers on its bottom • 5 medium-sized “Test” boxes,each with a different color label on its bottom. • A backpack, computer, or similar large object to hide the boxes behind Each group should choose one subject for this test. The other group member(s) will serve as tester and/or recorder(s). Place the backpack on the table between the subject and tester. Data should be recorded in the score sheets given to you by your instructor (please get one if you do not have one already). 1. Start with the medium-sized Standard box and the appropriate score sheet (Standard box: Medium-sized). On each trial, the tester should present both the Standard box and one Test box to the subject (the subject should notcontinue to hold on to the Standard). The tester should randomize both the order of the boxes handed out and which hand the standard is given to (to avoid handedness biases). 2. The subject should look at both boxes, lift both together (one in each hand), and indicate clearly which feels heavier. We will have to do many comparisons, sothe subject shouldperform only one lift of the two boxes in making each comparison (jiggling the boxes up and down is fine). “Equal” or “I don’t know” is not an acceptable answer—you will have to choose one or the other. This is why such tests are known as “2-alternative, forced choice(2-AFC)” behavioral tests in psychophysics. 3. The tester should show the chosen (judged heavier) box to the recorder. The recorder should record a “0” if the Standard box was judged to be heavier and a “1” if the Test box was judged to be heavier. The recorder’s answers should not be visible to the subject!4. Repeat with (approximately) randomized order of presentation of the Test boxes, as well as approximately randomized side to which the standard box is presented until each Test box has been tested against the Standard box exactly 10 times (i.e. until the sheet is completely filled with 50 total comparisons). 5. Repeat steps 1-4, using the appropriate score sheet, for the small- and large-sized Standard boxes. 6. For each Standard sized box, compute the probability (i.e., % of times) that the Test box was deemed to be heavier (or, equivalently, the standard box was judged lighter) for each test box color. 7. Enter your summary results (bottom row of each scoresheet), with percentages entered as numbers (e.g. “60” rather than “60%”)into the Google sheet posted on Canvas. Please make sure to enter your data in the row of the summary sheet corresponding to your section number and group letter.
9 PRELAB QUESTION 1: Estimating parameters of nonlinear functions by transforming variables. To analyze the data from the experiment described above, we will fit our threshold psychophysics behavioral data to a nonlinear function. By “fitting the data to a nonlinear function,” we meanestimating the unknown parameters of the nonlinear function from our data. Directly trying to fit the parameters of nonlinear functions is a tricky and error-prone procedure. Therefore, when possible, it is good practice to see if the function can be described alternatively as a linear function of some transformed version of the original variables. Below is a simple example to get you used to transforming variables and plotting functions both in the original variables and in the transformed variables. a) [2 points] Plot the function y = 3e-2xas a function of x for values of x ranging from x=0 to x=3. Notice that y is a nonlinear function of x. Feel free to draw this by hand –the point of this question is for you to get a feel for nonlinear and linear functions and how to change variables to convert nonlinear functions into linear functions of a transformed variable. b) [7 points] Define a new variable z = ln(y) where ln is the natural logarithm. (i) [3 points] Write an equation for this new variable z as a function of x by taking the logarithm of both sides of the original (y versus x) equation. (ii) [2 points] Plot z as a function of x. Feel free to draw this by hand. (iii) [3 points] The equation and plot should be in the form of a line. What is the slope of this line? What is the z-intercept of this line? In lab, we will be doing a similar type of transform of our data in order to be able to fitthe parameters of a function that describes our data. That is, in order to determine the best-fit parameters of the original, nonlinear function describing our data, we will first plot our data in terms of a transformed variable z so that z versus x is a linear function whose parameters can easily be fit by a linear trendline in Excel or other statistical analysis package.
10 8. For all group members: Present all 5 test boxes, the small Standard box, and the large Standard box to each group member, starting with the group member(s) who were not the subject in the previous part. Each group member should judge which color test box is closest in weight to the small Standard and which is closest in weight to the large Standard (note: you can also say “halfway between” or make similar gradations if helpful). For these judgments, note below which choice came from the group member who was the subject in the previous part and, if there was both a tester and a recorder, who was the recorder: The small Standard box: __________ , __________ , __________ The large Standard box: __________ , __________ , __________ 9. Weigh the test and standard boxes to see how close you were in your guesses. You can round to the nearest 5 grams. (Note: weigh all boxes even if some were never chosen, as you will need this information for the graphs in your postlab assignment.) D. Wrap-up Make sure you entered your summary data in the Google sheet posted on Canvas (see step 7 above) before leaving.