Neural Mechanisms underlying the planning of sequential saccades

Abstract

Saccades are rapid eye movements that we continually make (about 2-3 times per second) to look around and scan our visual environment. Though we effortlessly execute saccadic eye movements all the time, they are not just reflexive movements; saccades have been shown to involve multifaceted cognitive control mechanisms. This property of saccades, combined with the fact that saccadic parameters are easily measurable, and that the neural circuitry for saccade generation is fairly well established, has established saccadic eye movements to be an excellent tool to study motor planning and decision-making. However, much of the work done on saccade planning has been limited to understanding the production of single saccades. Natural behavior entails making multiple saccadic movements in a sequence to achieve day-to-day tasks such as reading a book. How are sequential saccades planned? This question forms the broad theme of this thesis.

The neural correlates of sequential saccade planning were scouted for in the macaque frontal eye field (FEF), a prefrontal area containing neuronal populations that undertake saccadic decision-making. Visual-salience neurons of the FEF have been shown to encode targets for upcoming saccades and movement-related neurons of the FEF have been shown to control the time of saccade initiation, providing a good link between neural activity and behaviorally measured reaction times. However, much of the neural underpinnings of saccade programming in the FEF have been uncovered using tasks involving single, isolated saccades. Motivated by this, I explored the mechanisms by which FEF neurons contributed to the programming of saccade sequences for this thesis, using single-unit electrophysiological recordings from the FEF of two macaques as they performed a sequential saccade task.

Sequential saccade programming can, in principle, operate through two major modes: serial or parallel. Behavioral measures, like short inter-saccadic intervals, strongly indicate that multiple saccade plans can proceed in parallel. However, direct neural evidence of parallel programming in the FEF neuronal population that strongly link to behavior, i.e. movement neurons, is lacking. First, I show that FEF movement-related activity can start ramping-up for the second saccade before the first saccade execution is complete, and much before visual feedback from the first saccade can reach FEF, thereby providing neural correlates of parallel programming of sequential saccades. Perceptual processing in the FEF has been shown to precede motor processing in visual search tasks, and consistent with that notion, FEF neurons with visual activity were also able to augment activity related to the second target whilst the first saccade plan was still underway.

After finding neural evidence of parallel programming, I characterized the limits of parallel programming. Numerous studies have shown that when two motor plans overlap closely, processing bottlenecks arise to inhibit the programming of the second plan, and is behaviorally manifested by the progressive lengthening of the second task reaction time, as the temporal gap between the two tasks decreases. This feature of increase in the second task latency has been observed in sequential saccade tasks as well. Neural correlates of processing bottlenecks were found in the responses of FEF movement neurons, wherein for the second saccade plan, the rate of the growth of activity was perturbed and the threshold of saccade initiation was increased, in a degree proportional to the level of concurrence of the two saccade plans. The locus of processing bottlenecks was found to be at the level of FEF movement-related neurons, whereas the activity of the visual neurons indicated that visual processing for perceptually simple tasks might constitute a pre-bottleneck stage. Evidence of activity perturbations was also found for the first saccade plan, supporting capacity-sharing theories of processing bottlenecks, as opposed to single-channel bottleneck theories which postulate that only the second plan is gated by inhibitory control while the first can pass unabated. Together, the results suggest that processing bottlenecks in sequential saccades originate in the partitioning of the brain’s limited processing ‘capacity’ by simultaneously active motor plans, due to which, inhibitory control is applied on both the first and second saccade plans, to prevent straining of the aforesaid capacity.

Finally, I have examined peripheral signatures of sequential saccade planning. Recent studies using single saccade paradigms have shown that the function of FEF as a center of cognitive control is not limited to saccade eye movements, but can be generalized to the control of eye-head gaze shifts. Rapid presaccadic recruitment of dorsal neck muscle activity has been shown to occur after FEF both with single-unit microstimulation and trans-cranial magnetic stimulation, even under head-restrained conditions where no overt head movement is being brought about by the neck muscles. To investigate whether such presaccadic recruitment occurs during sequential saccade planning or is gated out by inhibitory control, I recorded electromyographic (EMG) activity of motor units of the dorsal neck muscle as macaques performed the same sequential saccade task used for neural recordings. Neck muscle EMG showed leakage of FEF planning signals even for sequential saccades: peripheral correlates of parallel programming and processing bottlenecks were observed, with the activity mirroring that of FEF movement neurons. The correspondence of the results between the FEF and periphery suggest that a tight link exists between the eye and head systems, validating the hypothesis of a common gaze command originating in the FEF. The rapid recruitment of neck muscle activity observed for the second saccade before the completion of the first, also suggested that inhibitory control gates like basal ganglia do not preferentially intercept sequential saccade signals in the FEF-neck muscle circuit.

In summary, the results in this thesis provide direct neurophysiological evidence of behaviorally established features of sequential saccade planning such as parallel programming and processing bottlenecks. The fact that signatures of FEF movement responses can be captured at the level of the dorsal neck muscle suggests that the functional channel connecting FEF and the motor periphery is preserved even during sequential saccade planning and allows central responses to rapidly pass downstream by default, and perhaps prepare for an anticipated head movement in conjunction with the upcoming saccade. In cases where no head movement is elicited or where the head is restrained, inhibitory control mechanisms might come into play later and prevent supra-threshold rise of neck muscle activity.