My Graduate Research

All of my graduate work was completed in the beautiful Ashton Graybiel Spatial Orientation Lab, all thanks to my amazing advisors: Paul DiZio and James Lackner.

Project 1.1: Characterizing the learning of balance control in the absence of peripheral mechanisms

When I joined the lab, I had no known destiny and the turbulence and uncertainty of research obscured my vision of a clear path, and so sometimes I would stumble into bubbling cesspools of failure and sit there mired and confused. Lost within that labyrinth of failure, I met the Minotaur...it was the Beast of the Lab, known as the HULK and as the MART and as the MARS (the Multi Axis and Rotation System). I realized then that my salvation would only come by befriending this beast, by earning its trust and by majestically riding it out from the cloud of uncertainty. It took me years to become one with the beast...to understand how it moves and how it reacts when a human sits deep within it.

I began with the simplest experiment possible: strap blindfolded humans into the MARS and study how they learn to stabilize and balance and make peace with the beast.

The MARS is programmed to behave like an inverted pendulum, which means that just like balancing a pencil on your fingertip, the machine 'naturally' falls to the left or right if you do nothing. We chose inverted pendulum dynamics because it is a very simple way to approximate and simulate human postural balancing. To balance the machine, people used an attached joystick (eg, if a person is falling to the left, they will deflect the joystick to the right).

Usually when we balance in normal life, we rely on both central mechanisms (such as signals going from your brain that tell your leg muscles what to do) and peripheral mechanisms (such as reflexes and the biomechanical properties of your legs). Isolating and studying only the central mechanisms has traditionally been hard...after all, how can you study human balancing without using your legs? Thankfully the MARS allows us to do that because people are seated in the machine and use a joystick to control it. And so, my first project became: What do people learn when they have to balance without using their legs and reflexes?

As I began to look at the data, letting it enter in and out of my mind through conscious thought and dreams, I developed a deep desire to explore the vastness of the data....averages and standard deviations could not satisfy me...perhaps because of my youthful days of passionate and ambivalent romance with physics. So, I created an arsenal of mathematical metrics to quantify the learning, many of which were motivated from phase plots and the stabilogram diffusion function. You will have to read my papers to understand them better, for now, I leave you with a heart-throbbing figure below and if you are interested in how it turned into a series of art projects, watch my presentation (start at 8:40 for an intro of the paradigm, or at 12:50 for the data and then the artwork: https://livestream.com/accounts/3261852/events/8277165/videos/178138257

In general, we found that when people balance in the absence of peripheral mechanisms (the passive properties of the legs and the reflexes), they learn to minimize oscillations and avoid the crash boundaries by using smaller joystick deflections that are more intermittent. From metrics motivated from phase plots we found that they learned to reduce the number of destabilizing joystick deflections by better understanding the timing of when to make joystick movements (i.e. the phase relationship between position and velocity). We also found, to my surprise, that as people became better at controlling the machine they learned to let it fall for longer before changing its direction. To me this means that people learn to make peace with the beast. They learnt that balancing the machine, like life, isn't about responding to every small bump but instead is about being peaceful and waiting to make a small corrective movement at the right time.

In the future, we will make a small video on Mindfulness and the Machine, which will explore The Human's journey through the experiment, starting from panic, fright and anxiety and finishing in a womb-like nirvana state.

To obtain more details, check out my dissertation at the end of the page and my first paper below. Here is a video of my early work that I presented to the Brandeis biology and neuroscience community: https://www.youtube.com/watch?v=O2Kz10Jc_OU

My first paper: Vimal, Vivekanand Pandey, James R. Lackner, and Paul DiZio. "Learning dynamic control of body roll orientation." Experimental brain research 234.2 (2016): 483-492.

Project 1.2: Characterizing the learning of balance control in the absence of peripheral mechanisms and gravitational cues (Roll Plane Rotations)

The core part of my dissertation was focused on examining what people learn when balancing in the absence of relevant gravitational cues. Answers to this question are useful in understanding vestibular diseases and the causes of spatial disorientation in astronauts and military pilots.

But before I continue...one big question is: How on Earth could I ever get rid of relevant gravitational cues?

To understand this, I will first review the vestibular system which is our 'inner sense of balance' and is found in our inner ears (for a more thorough explanation watch the second youtube video in the Introduction). The vestibular system consist of semicircular canals and the otolith organs. The canals are made from 3 semicircular tubes that are filled with fluid and each of the tubes are 90 degrees from the others. There are little sensory hairs that are suspended in the fluid and when you rotate, the fluid remains mostly still as your skull moves, causing the hairs to deflect and send down a signal telling you that you are moving. As a result, the semicircular canals give motion information (acceleration and velocity) when a person is rotating.

The otolith organs (pictured below) are these fine little hairs suspended in gelatinous material and adorned by crystals and they detect linear acceleration. For my project, the most important thing that they detect is tilt from gravitational vertical. For example, if you tilt 10 deg from upright (the gravitational vertical), the force of gravity pulls on the crystals which deflects the sensory hairs, sending a signal saying that you are tilted.

In the Upright Roll Condition (discussed in Project 1.1 and pictured below), I strap blindfolded people into the MARS (the multi-axis rotation system) which is programmed to behave like an inverted pendulum. If you do nothing with the attached joystick, the MARS 'naturally' falls to the left or right (Roll rotations are where you are falling to the left or right, Pitch Rotations are when you are going forward and backwards and Yaw Rotations are when you are twisting). Because participants are blindfolded and wear noise cancelling headphones, they mostly rely on their vestibular and somatosensory (the touch receptors on your skin) systems to inform them about where they are. Therefore, in the Upright Roll Condition, the sensory information they receive, consists of:

• Otoliths: information about their position (the more they tilt from the upright, the more the hair deflects, telling them where they are).

• Canals: information about their motion

• Somatosensory system: both position and motion

Now I can answer the question from before: How on Earth can I ever get rid of relevant gravitational cues? The answer is that I put people on their back and have the MARS behave just like before (if you do nothing, you fall to the left or right and you have to use a joystick to keep yourself stabilized around the balance point). In the Back Roll Condition (pictured below), people cannot use gravitational information to determine how much they are tilted. Let me explain that. When you are on your back, your otolith hairs just point down, telling you that you are on your back. Now if you tilt to the left, the otolith hairs still just say that you are on your back. Because you are always 90 degrees from the vertical and because you no longer tilt relative to upright, the otoliths cannot give you any useful information about how much you are tilted. In other words, you cannot use gravitational information to tell you how much you are tilted from the balance point. To summarize, the sensory information you receive:

• Otoliths: NO useful information about a person's angular position (because they are always 90deg from upright).

• Canals: information about their motion

• Somatosensory system: only information about motion

So what happened? Based on prior literature, one would guess that people in the Back Roll Condition would perhaps perform a little worse but nevertheless should be able to do the task because within the brain there are neural integrators that can take velocity and make estimates of position. Surprisingly, we found that people were really very bad at balancing in the Back Condition. Across majority of metrics, they collectively showed no learning. In fact, 90% of them reported that they had no idea where they were, some people said that they felt like they were spinning 360deg even though the machine could only move 60deg to the left or right. Other people reported feeling like they were completely upside down even though they were always on their back.

Therefore, in this condition, collectively, people showed very minimal learning and they reported spatial disorientation. Even more interesting is that they all had a very characteristic pattern of positional drifting, where they would balance and slowly drift farther and farther away without realizing it.

The first graph, below, shows a person in the final trial of the Upright Roll Condition. The purple line represents the angular position and you can see that she has mastered the task, minimally oscillating around the balance point (0 deg). In contrast, the same person in the final trial of the Back Roll Condition shows positional drift.

Before I describe the meaning, implications and future directions of this research, I need to first discuss Project 1.3 (below).

For more details, read my second paper below: Vimal, Vivekanand Pandey, Paul DiZio, and James R. Lackner. "Learning dynamic balancing in the roll plane with and without gravitational cues." Experimental brain research235.11 (2017): 3495-3503.

Project 1.3: Characterizing the learning of balance control in the absence of peripheral mechanisms and gravitational cues (Yaw Axis Rotations)

The results of Project 1.2 suggest that the reason why people performed so poorly in the Back Roll Condition is because of the absence of relevant gravitational cues, which is something essential for proper balance control. However, some people could argue against this idea by proposing that the poor performance and minimal learning was caused by being oriented on the back, because perhaps the vestibular system is not designed to work in that orientation. To answer this doubt, I ran people in the Upright Yaw Condition. In this condition, the chair was in the upright orientation and was programmed to behave like an inverted pendulum, just like the Roll Condition (i.e. if you do nothing, the MARS either rotates clockwise or counterclockwise). Similar to the Roll Condition experiments, people were told to use the joystick to stabilize themselves around the balance point. In these rotations, because they are always parallel to the gravitational vertical, they do not receive any useful information from gravitational cues about how far they were from the balance point (because you don't tilt relative to it). As a result, the Upright Yaw Condition is very similar to the Back Roll Condition, the only difference being that it is in the upright orientation, which is where the vesibular system is designed to perform optimally. I compared the Upright Yaw group to the Back Yaw group, where people were on their back, twisting either clockwise or counterclockwise. The Back Yaw Condition is similar to the Upright Roll condition because in both situations the head tilts relative to the gravitational vertical.

Long story short, the results confirmed the findings from Project 1.2: In the absence of relevant gravitational cues (where people can only rely on motion information), there is very minimal learning, poor performance and a characteristic pattern of positional drifting.

This is surprising because prior work reported that people were fairly accurate when they were passively rotated while blindfolded in the Upright Yaw axis and then asked to report how far they were displaced. What is different about my experiments? The main difference is that prior work only rotated the people once, whereas my active balancing task requires people to make dozens of movements within a trial.

There are two possibilities that can explain why multiple rotations can lead to such problems. The first is that the neural integrator in the brain that takes velocity information and then estimates position (e.g. distance=velocity*time) is not adequate for dynamic stabilization and accumulates error in each movement. The other possibility is that as a movement begins, there is a portion of the movement that is below the detection threshold and therefore it is not included in the integration when a position estimate is made. Projects 2.3 and 2.4, in the Current Research section, explore these ideas further.

• • • My third paper: Vimal, Vivekanand Pandey, James R. Lackner, and Paul DiZio. "Learning dynamic control of body yaw orientation." Experimental brain research (2018): 1-10.

Project 1.4: Elucidating the two dissociable components of balance control

Initial analysis of Project 1.2's data showed that people in the Back Roll Condition were unable to collectively decrease the variability of their movements because they could not use relevant gravitational cues. Not only that, some people actually got worse as time went by! Therefore, looking at just averages and standard deviations, I would quickly conclude that subjects in the Back Roll Condition learned nothing........but I so desperately wanted to escape the mundane and limited world of averages and standard deviations. I knew that there was an entire universe of depth and beauty in the data...I could sense that there was some unexplored Truth hidden within. So, I loaded my entire arsenal of mathematical metrics that I created in Project 1.1 and went on an adventure to search my data for beautiful patterns........and I found them!

I discovered that every single person on the Back Condition learned to decrease the number of Crashes (the number of times they hit the boundary) and the percentage of destabilizing joystick deflections. This included even those people who's motion became wilder with time and who's performance, from the perspective of averages and standard deviations, got worse with time. So what could this mean? This meant that in the absence of a relevant gravitational cues, there was something that people learned which suggested a hidden process of balancing that has nothing to do with aligning with gravity.

This idea was further supported by experiments done in Project 1.2. In these experiments, I had one group of people balance in the Upright Roll Condition (they have relevant gravitational cues) on the first day and then in the Back Roll Condition(they do not have relevant gravitational cues) on the second day. I originally thought that the first day of practice in the Upright would lead to better performance on the second day. But to my surprise, there was no transfer at all. However, when I had a different group of people first balance in the Back Roll Condition and then, on the second day, they balanced in the Upright Roll Condition, I found significant transfer and enhanced early learning. We hypothesized that the reason there was no transfer from the Upright to the Back was because the Upright is dominated by the task of aligning to the gravitational vertical, however the reason why something transferred from the Back to the Upright is because people learned about dynamic stabilization independent of gravity when on the Back, and this is also shared in the Upright.

Therefore, our work suggests that there are 2 dissociable components of balance control:

1. Alignment to the Gravitational Vertical. This relies mostly on gravitational cues sensed by the otoliths and touch receptors on the skin. This dominates most of the balancing we experience in normal life.

2. Dynamic Stabilization: This relies mostly on motion cues from the canals and touch receptors on the skin. This is the only type of control you can do when in the Back Roll Condition or the Upright Yaw Condition.

Project 2.1 in the Current Research section uses this finding to develop a novel training program to enhance performance in our spaceflight analog condition (Back Roll Condition).

A detailed overview of my graduate research can be found in my dissertation provided below: