Beyond the Paper: A Conversation with Omri Nachmani and Gunnar Blohm

Interviewed by Rosalind S.E. Carney, DPhil

Registered Reports are an empirical article in which the methods and proposed analyses are preregistered and reviewed prior to research being conducted. High-quality protocols are then provisionally accepted for publication before data collection commences.

For humans to see an object clearly, its image must fall on a highly specialized region in the retina called the fovea. Foveal tracking of moving objects in the environment is accomplished by smooth pursuit eye movements. However, due to visual feedback delay and sensorimotor noise, the eye can progressively lag behind the target. When significant position error is accumulated, a catch-up saccade may be triggered to re-foveate the target. However, the mechanistic explanation for how the trigger decision is made by the brain has eluded researchers for decades. Nachmani and colleagues present behavioral data in human subjects complemented by computational model predictions in support of an intuitive trigger mechanism that relies on a probabilistic estimation of future position error. These results add support for a common and shared sensorimotor process for saccades and smooth pursuit. Furthermore, by linking motor control to statistical decision-making, the authors offer a novel perspective on how sensorimotor prediction and uncertainty modulate oculomotor tracking behavior.

(left) Omri Nachmani, MSc, former graduate student, Center for Neuroscience Studies, Queen’s University, Ontario, Canada
(Right) Gunnar Blohm, Professor, Department of Biomedical and Molecular Sciences, Queen’s University School of Medicine, Ontario, Canada

RC: How did you become interested in this particular line of research? 

Gunnar: “As a PhD student, I was interested in the coordination between different types of eye movements, including saccades and smooth pursuit. I was a co-author of the de Brouwer et al. (2002) paper, which demonstrated that catch-up saccades were less likely to occur when the visual target re-crosses the fovea within 40–180 milliseconds. The 2002 publication provided a descriptive analysis of the behavioral interactions between the saccadic and smooth pursuit systems; however, a mechanistic explanation of coordination between the systems was lacking. Traditionally, eye movement modeling has been investigated in a systems approach in which all signals are deterministic. Only recently, partially inspired by the arm movement field literature, have researchers started to incorporate variability and stochasticity in their way of thinking. Recently, we proposed a stochastic decision model for saccade triggering during visual tracking (Coutinho et al., 2018) that relies on a probabilistic estimate of predicted position error. Informed by that decision model, we next hypothesized that saccade trigger time length and variability would increase when pre-saccadic predicted errors are small or visual uncertainty is high (e.g., for blurred targets), which we addressed in this paper. Therefore, a long build-up evolved our way of thinking about eye movement that culminated in this study.”

Omri: “I became interested in neuroscience during the third year of my undergraduate degree. I gravitated towards computational neuroscience because of the potential to simulate brain processes and behaviors. Testing computational stimulations is appealing to me as I can simultaneously gain valuable career skills, such as programming and statistical analysis, while exploring how the brain functions. When I joined Gunnar’s lab, I was presented with diverse options for projects. I chose to investigate how smooth pursuit and catch-up saccades contribute to the visual tracking of moving stimuli.”

How did you develop the methodology for the study?

Omri: “The methodology was largely guided by the previous work of the de Brouwer et al. (2002) paper. We added a few variables that we wanted to explore. For example, we investigated whether an increase in uncertainty of the position of the target would influence decision making and motor coordination to trigger a catch-up saccade or whether eye movement would remain in smooth pursuit. To do that, we used a blurred target, for which uncertainty of the target location is increased, to examine the influence of additional sensory noise on behavior. Each subject was seated in front of a large screen—the target was initially located at the periphery. Without any warning, the target moved centrally with a random velocity. The first part of the task was used to induce sustained smooth pursuit visual tracking by the subject. The second part of the task included an unpredictable step in the target position and a random change in target velocity. The latter part of the task is how we manipulated the parameters that feed into the catch-up saccade decision system. We modulated the target position and velocity, and the subject’s uncertainty of the target location, to determine if and when the brain decides to trigger a catch-up saccade.”

Gunnar: “Humans are fairly good at visually tracking moving objects, therefore, it is hard to really study what the decision mechanism is for the triggering of catch-up saccades. By artificially throwing people off by jumping the target to a different place, speeding it up or slowing it down, we essentially forced the brain to react to the movements of the image to determine how decision making occurs.”

What were the main findings in this eNeuro paper?

Omri: “The central finding of the paper is that the brain continually tries to predict events of interest in the environment to react efficiently. In visual tracking, when the prediction of the target location does not match the incoming stimulus, the brain reacts promptly to resolve the mismatch between the expected and actual position of a target. We found that the actual reaction time is modulated by a predicted error-based decision process, in which large predicted errors are significantly more likely to trigger catch-up saccades than small predicted errors or errors with low uncertainty. These findings are relevant to the visual tracking of any moving object, such as when we are driving or playing sports.”

Why is this eNeuro paper an important contribution to the field?

Gunnar: “Since at least the 1960s, researchers have tried to determine how different eye movement motor systems, such as the smooth pursuit and saccadic systems interact to achieve the same goal—visual tracking. Within the past two decades, research has shown that smooth pursuit and saccadic systems have partially overlapping neural substrates, with the suggestion that these two types of eye movements may originate from a single sensorimotor process. In our paper, we propose, for the first time, a mechanistic answer to how the brain integrates smooth pursuit and saccadic eye movements to track a visual target. Our results support the hypothesis that the inputs to both eye movement systems are intertwined and are used in combination to compute predictions of future position errors and that these predictions are modulated by sensory uncertainty. What is particularly interesting is that another recent study on limb movement has suggested an optimal feedback control framework, which includes sensory prediction as a component (Crevecoeur and Gevers, 2019). Therefore, we are now learning more about how the brain uses predictive mechanisms for motor adaptation processes.”

How was your experience publishing your paper as an eNeuro Registered Report?

Omri: “Gunnar’s lab had not submitted a pre-registered study before, so we were initially uncertain how the process works until I reviewed the eNeuro guidelines. I was a little nervous being the first person in the lab to submit a Registered Report, but in the end, I was very glad that I did so. Pre-registration meant that I had an in-principle acceptance of a publication very early in my graduate studies. Receiving the reviewers’ feedback gave me both validation and a safety net to be able to conduct my experiments without any anxiety that our approach contained flaws or inadequate experimental conditions.”

Gunnar: “It was a good experience, and I would submit a Registered Report again. I like that the pre-registration process forces you to generate your hypotheses before results and be very specific about the research plan. In terms of statistical analyses, there is a clear benefit of having the proposed methods reviewed before gathering data. In terms of the methodology, the traditional way of paper submission carries the stress of whether the reviewers will not like the task that was used or ask for other control conditions. With pre-registration, important methodological considerations are easy to incorporate before data collection, thus lowering the rework associated with the traditional peer review process.”

Was the experimental plan improved by the pre-registration process?

Omri: “The reviewers had very good feedback about how to analyze the data. Receiving such feedback so early in the project was instrumental in how the paper turned out. For example, the reviewers suggested that we incorporate Bayesian statistics to complement the frequentist descriptions. In addition, the reviewers commented on the number of bins in our ANOVA analyses to eliminate distortion of the data. Another recommendation was to increase the number of participants to ensure adequate effect size. Based on the reviewers’ comments, we refined and changed our preliminary research design to maximize the outcome of the research.”

What consideration(s) would you convey to other researchers about a Registered Report submission?

Gunnar: “The one thing that might be interesting to point out to the readership is that, in general, many people think that there are additional work constraints or a longer process involved in pre-registrations. But ultimately, I think the pre-registration process leads to frontloading a portion of the work that needs to be done in any case. The initial reviews are an additional scientific check so that you do not have to revise and change your methodology post hoc. In that sense, pre-registration can potentially save time and money in terms of running experiments. When we received the in-principle acceptance, we knew that we could go forward without any further delays in terms of publishing, as long as we stuck to the approved data collection and data analysis paradigms. From a supervisor’s perspective, I thought that the pre-registration process was particularly beneficial to a trainee because the timing of the literature review and the experimental design coincided. Often, trainees start to collect data, and a thorough literature review is performed when the manuscript is being written. In this study, Omri was deeply immersed in the relevant literature from the start, which aided the research design. In particular, for studies that include human participants, approval of the research design before experimentation eliminated the concern that we would have to ask the original participants to return for more experiments.”

Omri: “For students, a Registered Report may align better with the timeline of a doctoral program than a master’s program. On the flip side, all the front loading of the literature review and experimental design considerably reduced my workload near the end of my program which is typically a very stressful time. Once I finished the data analysis, I only had to write the Results and Discussion of the manuscript because the rest was already completed.”

How would you describe the mentoring philosophies in the lab?

Gunnar: “I am a strong advocate for open science. However, with open science, you need to be able to set up the data recording and data analysis programs in a way that other researchers can make sense of it—otherwise you are sharing code that is gibberish. Nowadays, I tell all my trainees that everything they produce will be published, so it is crucial to organize the data appropriately. The result is that trainees master different ways of managing code and data, in addition to managing the project itself. In addition to being beneficial for science, these organization skills will be useful along the career path.”

Omri: “I admire Gunnar’s advocacy for open science. I like the idea of sharing resources and data within the scientific community to accelerate our common goal of further understanding brain function. In the lab, Gunnar allows his students to have a lot of hands-on experience without micromanaging. This approach gives the trainees enough time for their scientific exploration and problem-solving development, although Gunnar is happy to step in and give guidance in a collaborative process.”

What are your future research goals?

Omri: “I am currently in medical school at McMaster University, Ontario, Canada. I intend to stay in academia, so I am now looking for clinical research projects in neurology or neurosurgery.”

Gunnar: “In general, my lab is interested in understanding sensory-motor control, including aspects such as decision making, movement planning, and movement control. We want to develop further our modeling approach to gain insight into how individual neurons in the brain can achieve these tasks, for example, examining spiking neural networks. Current machine learning advances allow us to do experiments that were not previously possible; therefore, we will revisit aspects of sensory-motor control with contemporary approaches.”

References:

1. de Brouwer S, Yuksel D, Blohm G, Missal M, Lefèvre P (2002) What triggers catch-up saccades during visual tracking? J Neurophysiol 87:1646–1650.
2. Coutinho JD, Lefèvre P, Blohm G (2018) Confidence in predicted position error explains saccadic decisions during pursuit. bioRxiv 396788.
3. Crevecoeur F, Gevers M. (2019) Filtering Compensation for Delays and Prediction Errors during Sensorimotor Control. Neural Comput. 31(4):738–64.

Read the full article:

Predicted Position Error Triggers Catch-Up Saccades during Sustained Smooth Pursuit
Omri Nachmani, Jonathan Coutinho, Aarlenne Z. Khan, Philippe Lefèvre and Gunnar Blohm