A dynamic, imperturbable link between midbrain activity and saccade velocity

Abstract

We make a saccadic eye movement once every few hundred milliseconds; however, the neural control of saccade execution is not fully understood. Dynamic, moment-by-moment variations in saccade velocity are typically thought to be controlled by neurons in the lower, but not the upper regions of the brainstem. In a recent report, Smalianchuk et al. (Smalianchuk I, Jagadisan UK, Gandhi NJ. J Neurosci 38: 10156–10167, 2018) provided strong evidence for a role of the superior colliculus, a midbrain structure, in the instantaneous control of saccade velocity, suggesting the revision of long-standing models of oculomotor control.

Our eyes dart rapidly about the visual scene every few hundred milliseconds. These quick, jerk-like movements of the eyes, dubbed “saccades” in the late 1800s (Javal 1879), have long-been used as a model system for studying the neurobiological bases of perception, cognition, and action. An outstanding goal of systems neuroscience is to elucidate how the brain controls a saccade’s most defining characteristic — its dynamic, high-velocity profile. In a recent article published in The Journal of Neuroscience, Smalianchuk et al. (2018) demonstrate an imperturbable link between midbrain activity and saccade velocity that exists throughout the entire duration of the eye movement.

It is well established that the superior colliculus (SC), located in the midbrain, is heavily involved in the motor planning and execution of saccadic eye movements (Sparks 1986). Neurons in the intermediate and deep layers of the SC fire bursts of action potentials shortly before saccade onset, and artificial microstimulation of these neurons, when of sufficient intensity, evokes a saccade (Robinson 1972; Schiller and Koerner 1971; Wurtz and Goldberg 1971). The vector of an impending saccade is represented in the form of a place code within the SC’s topographically organized motor map; that is, the locus of SC activation determines the amplitude and direction of a saccade when triggered (Gandhi and Katnani 2011; Robinson 1972; Sparks 1986). In contrast, the overall level of SC activity has a causal influence on a saccade’s peak velocity, independent from its amplitude (Lee et al. 1988; Stanford et al. 1996). New findings by Smalianchuk et al. (2018) show that the relationship between SC activity and saccade peak velocity is only the tip of the iceberg.

The authors recorded the eye position and SC activity of head-fixed monkeys performing a visually guided saccade task. They correlated measures of SC activity with saccade velocity at each time point spanning the eye movement. The authors repeated this calculation over and over, systematically shifting the temporal delay between the two variables correlated on each iteration. Importantly, the computations were performed on amplitude-matched saccades, thus controlling for amplitude-dependent modulation of SC activity and saccade velocity (based on a well-documented relationship called the saccadic “main sequence”; Bahill et al. 1975). The analysis revealed two key findings. First, it produced a highly precise estimate of the efferent delay between SC activity and saccade kinematics for normal saccades: 12 ms. And second, it demonstrated that SC neuronal activity was strongly correlated with the velocity of a saccadic eye movement on a millisecond-by-millisecond basis — SC activity levels predicted not only saccade peak velocity, but also variability around the entire velocity profile.

Smalianchuk et al. (2018) then tested the fortitude of the correlation between SC activity and instantaneous saccade velocity. They perturbed saccade velocity by briefly puffing air into the eyes of the subject around the time of saccade execution and looked for concomitant changes in SC activity. Before reviewing their findings, it is worth noting that previously observed correlations between SC activity and motor control parameters have been exposed as superficial using this technique. That is, SC activity has been demonstrated to correlate linearly with the dynamic motor error of normal saccades; however, this relationship breaks down when saccades with perturbed trajectories are evoked by air puffed into the eyes (Goossens and Van Opstal 2000; Waitzman et al. 1991). Smalianchuk et al. (2018), on the other hand, discovered that the correlation between SC activity and saccade velocity persisted throughout the duration of the perturbed saccades.

Together, the findings by Smalianchuk et al. (2018) strongly suggest that the SC plays an essential role in controlling instantaneous saccade velocity, a role that was traditionally attributed to excitatory burst neurons within the lower brainstem, downstream from the SC (Van Gisbergen et al. 1981). The authors presented a plausible circuit-level framework that may account for their observations, while keeping in accordance with previous findings. The new framework may also help explain why the field has had difficulty finding two crucial components of classical theories of oculomotor control, i.e., the resettable neural integrator, and the comparator (Robinson 1975) — they may simply not exist. That is, classical theories posited that excitatory burst neurons (EBNs) are driven by a dynamic motor error signal, computed (via the neural comparator) by subtracting the current eye displacement (calculated by integrating velocity signals from the EBNs) from the desired eye displacement within a local feedback loop downstream from the SC (Robinson 1975). In contrast to this traditional model, the authors proposed a framework in which instantaneous saccade velocity signals, sent from the SC, directly drive the EBNs (rather than a motor error signal), thus obviating the need for a neural integrator and comparator, which are needed to compute a dynamic motor error signal. Therefore, the framework presented by the authors incorporates a new role for the SC in the instantaneous control of saccade velocity and it provides a parsimonious interpretation of the neurobiological circuitry that governs the dynamics of a saccade in flight.

Variability in the velocity of amplitude-matched saccades (i.e., saccade vigor), which the authors demonstrate correlates with that of SC activity (Smalianchuk et al. 2018), may arise from external inputs to the proposed neural circuit. One candidate region that may influence saccade vigor via its direct and/or indirect projections to the deeper layers of the SC is the frontal eye field (FEF; Leichnetz et al. 1981). Indeed, FEF activity has been demonstrated to correlate with and causally affect amplitude-independent changes in saccade peak velocity (Glaser et al. 2016; Kimmel and Moore 2007). Interestingly, a variety of cognitive processes (e.g., expectation of reward, decision making) have been shown to modulate FEF motor activity, SC visuomotor activity, and saccade vigor (Glaser et al. 2016; Horwitz and Newsome 2001; Ikeda and Hikosaka 2007; Seideman et al. 2018; Shadmehr et al. 2019; Stanford et al. 2010). Together, these empirical findings, along with recent theoretical work (Seideman et al. 2018), suggest that this cortico-subcortical pathway may help mediate the influence of cognitive signals on saccade vigor. A similar line of evidence suggests that the basal ganglia also play a crucial role in this transformation (Hikosaka and Wurtz 1983; Shadmehr et al. 2019). There is substantial interest in using saccade velocity as a behavioral marker to aid in the diagnosis and general understanding of cognitive/neurological disorders, especially those thought to be related to alterations in activity within these regions that project to the SC (e.g., attention-deficit/hyperactivity disorder, Huntington’s disease, and Parkinson’s disease; Munoz et al. 2003; Willard and Lueck 2014). Smalianchuk et al.’s (2018) findings are centrally linked and may one day prove valuable to these important efforts.

Learning how the brain controls saccade velocity might, in turn, advance our understanding of the neural bases of decision making. For example, one of the aforementioned studies found that, during performance of an urgent perceptual decision-making task (in which the time available for responding is limited), the probability of making a correct discrimination and the peak velocities of saccades used to report the subjects’ choices demonstrated parallel modulation profiles (Seideman et al. 2018). Furthermore, a preexisting model that was consistent with FEF activity recorded during the task could simultaneously replicate standard performance metrics (e.g., response time distributions, choice accuracy) as well as the observed modulations of saccade peak velocity. These results suggest that the observed variations in saccade peak velocity are derived from the very same neural activity that determines the direction, timing, and accuracy of saccadic choices. In fact, saccade peak velocity and simulated FEF activity both correlated with the statistical definition of decision confidence (Hangya et al. 2016) (SC activity is also implicated in computing decision confidence; Grimaldi et al. 2015) — reinforcing the idea that, under time pressure, decision-related computations and saccade velocity share a common neural code at the level of the cortex. Further empirical evidence is needed to confirm these model predictions and to elucidate whether and how kinematic/confidence signals are transformed between the FEF and SC.

Smalianchuk et al.’s (2018) results demonstrate that the primary circuitry responsible for controlling instantaneous saccade velocity extends to a level of the brain that is higher than previously thought — i.e., to the SC, a nexus of perceptual, cognitive, and motor signals in the midbrain. The neural control of saccade velocity is a topic of widespread relevance, and these new findings by Smalianchuk et al. (2018) provide a strong foundation upon which a broad range of future investigations may build.

GRANTS

This work was supported by the National Institutes of Health through National Eye Institute grant F31EY029154 awarded to J. A. Seideman.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author.

AUTHOR CONTRIBUTIONS

J.A.S. drafted manuscript; J.A.S. edited and revised manuscript; J.A.S. approved final version of manuscript.