Quantifying effects of stochasticity in reference frame transformations on posterior distributions

Motor "laziness" constrains fixation selection in real-world tasks

Humans coordinate their eye, head, and body movements to gather information from a dynamic environment while maximizing reward and minimizing biomechanical and energetic costs. However, such natural behavior is not possible in traditional experiments employing head/body restraints and artificial, static stimuli. Therefore, it is unclear to what extent mechanisms of fixation selection discovered in lab studies, such as inhibition-of-return (IOR), influence everyday behavior. To address this gap, participants performed nine real-world tasks, including driving, visually searching for an item, and building a Lego set, while wearing a mobile eye tracker (169 recordings; 26.6 h). Surprisingly, in all tasks, participants most often returned to what they just viewed and saccade latencies were shorter preceding return than forward saccades, i.e., consistent with facilitation, rather than inhibition, of return. We hypothesize that conservation of eye and head motor effort ("laziness") contributes. Correspondingly, we observed center biases in fixation position and duration relative to the head's orientation. A model that generates scanpaths by randomly sampling these distributions reproduced all return phenomena we observed, including distinct 3-fixation sequences for forward versus return saccades. After controlling for orbital eccentricity, one task (building a Lego set) showed evidence for IOR. This, along with small discrepancies between model and data, indicates that the brain balances minimization of motor costs with maximization of rewards (e.g., accomplished by IOR and other mechanisms) and that the optimal balance varies according to task demands. Supporting this account, the orbital range of motion used in each task traded off lawfully with fixation duration.

Integration of landmark and saccade target signals in macaque frontal cortex visual responses

Visual landmarks influence spatial cognition and behavior, but their influence on visual codes for action is poorly understood. Here, we test landmark influence on the visual response to saccade targets recorded from 312 frontal and 256 supplementary eye field neurons in rhesus macaques. Visual response fields are characterized by recording neural responses to various target-landmark combinations, and then we test against several candidate spatial models. Overall, frontal/supplementary eye fields response fields preferentially code either saccade targets (40%/40%) or landmarks (30%/4.5%) in gaze fixation-centered coordinates, but most cells show multiplexed target-landmark coding within intermediate reference frames (between fixation-centered and landmark-centered). Further, these coding schemes interact: neurons with near-equal target and landmark coding show the biggest shift from fixation-centered toward landmark-centered target coding. These data show that landmark information is preserved and influences target coding in prefrontal visual responses, likely to stabilize movement goals in the presence of noisy egocentric signals.

Effector-dependent stochastic reference frame transformations alter decision-making

Psychophysical, motor control, and modeling studies have revealed that sensorimotor reference frame transformations (RFTs) add variability to transformed signals. For perceptual decision-making, this phenomenon could decrease the fidelity of a decision signal's representation or alternatively improve its processing through stochastic facilitation. We investigated these two hypotheses under various sensorimotor RFT constraints. Participants performed a time-limited, forced-choice motion discrimination task under eight combinations of head roll and/or stimulus rotation while responding either with a saccade or button press. This paradigm, together with the use of a decision model, allowed us to parameterize and correlate perceptual decision behavior with eye-, head-, and shoulder-centered sensory and motor reference frames. Misalignments between sensory and motor reference frames produced systematic changes in reaction time and response accuracy. For some conditions, these changes were consistent with a degradation of motion evidence commensurate with a decrease in stimulus strength in our model framework. Differences in participant performance were explained by a continuum of eye–head–shoulder representations of accumulated motion evidence, with an eye-centered bias during saccades and a shoulder-centered bias during button presses. In addition, we observed evidence for stochastic facilitation during head-rolled conditions (i.e., head roll resulted in faster, more accurate decisions in oblique motion for a given stimulus–response misalignment). We show that perceptual decision-making and stochastic RFTs are inseparable within the present context. We show that by simply rolling one's head, perceptual decision-making is altered in a way that is predicted by stochastic RFTs.

Integration of allocentric and egocentric visual information in a convolutional/multilayer perceptron network model of goal-directed gaze shifts

Allocentric (landmark-centered) and egocentric (eye-centered) visual codes are fundamental for spatial cognition, navigation, and goal-directed movement. Neuroimaging and neurophysiology suggest these codes are initially segregated, but then reintegrated in frontal cortex for movement control. We created and validated a theoretical framework for this process using physiologically constrained inputs and outputs. To implement a general framework, we integrated a convolutional neural network (CNN) of the visual system with a multilayer perceptron (MLP) model of the sensorimotor transformation. The network was trained on a task where a landmark shifted relative to the saccade target. These visual parameters were input to the CNN, the CNN output and initial gaze position to the MLP, and a decoder transformed MLP output into saccade vectors. Decoded saccade output replicated idealized training sets with various allocentric weightings and actual monkey data where the landmark shift had a partial influence (R² = 0.8). Furthermore, MLP output units accurately simulated prefrontal response field shifts recorded from monkeys during the same paradigm. In summary, our model replicated both the general properties of the visuomotor transformations for gaze and specific experimental results obtained during allocentric–egocentric integration, suggesting it can provide a general framework for understanding these and other complex visuomotor behaviors.

Spatiotemporal transformations for gaze control

Sensorimotor transformations require spatiotemporal coordination of signals, that is, through both time and space. For example, the gaze control system employs signals that are time-locked to various sensorimotor events, but the spatial content of these signals is difficult to assess during ordinary gaze shifts. In this review, we describe the various models and methods that have been devised to test this question, and their limitations. We then describe a new method that can (a) simultaneously test between all of these models during natural, head-unrestrained conditions, and (b) track the evolving spatial continuum from target (T) to future gaze coding (G, including errors) through time. We then summarize some applications of this technique, comparing spatiotemporal coding in the primate frontal eye field (FEF) and superior colliculus (SC). The results confirm that these areas preferentially encode eye-centered, effector-independent parameters, and show-for the first time in ordinary gaze shifts-a spatial transformation between visual and motor responses from T to G coding. We introduce a new set of spatial models (T-G continuum) that revealed task-dependent timing of this transformation: progressive during a memory delay between vision and action, and almost immediate without such a delay. We synthesize the results from our studies and supplement it with previous knowledge of anatomy and physiology to propose a conceptual model where cumulative transformation noise is realized as inaccuracies in gaze behavior. We conclude that the spatiotemporal transformation for gaze is both local (observed within and across neurons in a given area) and distributed (with common signals shared across remote but interconnected structures).

Reaching around obstacles accounts for uncertainty in coordinate transformations

When reaching to a visual target, humans need to transform the spatial target representation into the coordinate system of their moving arm. It has been shown that increased uncertainty in such coordinate transformations, for instance, when the head is rolled toward one shoulder, leads to higher movement variability and influence movement decisions. However, it is unknown whether the brain incorporates such added variability in planning and executing movements. We designed an obstacle avoidance task in which participants had to reach with or without visual feedback of the hand to a visual target while avoiding collisions with an obstacle. We varied coordinate transformation uncertainty by varying head roll (straight, 30° clockwise, and 30° counterclockwise). In agreement with previous studies, we observed that the reaching variability increased when the head was tilted. Indeed, head roll did not influence the number of collisions during reaching compared with the head-straight condition, but it did systematically change the obstacle avoidance behavior. Participants changed the preferred direction of passing the obstacle and increased the safety margins indicated by stronger movement curvature. These results suggest that the brain takes the added movement variability during head roll into account and compensates for it by adjusting the reaching trajectories.NEW & NOTEWORTHY We show that changing body geometry such as head roll results in compensatory reaching behaviors around obstacles. Specifically, we observed head roll causes changed preferred movement direction and increased trajectory curvature. As has been shown before, head roll increases movement variability due to stochastic coordinate transformations. Thus these results provide evidence that the brain must consider the added movement variability caused by coordinate transformations for accurate reach movements.

Saccade-induced changes in ocular torsion reveal predictive orientation perception

Natural orienting of gaze often results in a retinal image that is rotated relative to space due to ocular torsion. However, we perceive neither this rotation nor a moving world despite visual rotational motion on the retina. This perceptual stability is often attributed to the phenomenon known as predictive remapping, but the current remapping literature ignores this torsional component. In addition, studies often simply measure remapping across either space or features (e.g., orientation) but in natural circumstances, both components are bound together for stable perception. One natural circumstance in which the perceptual system must account for the current and future eye orientation to correctly interpret the orientation of external stimuli occurs during movements to or from oblique eye orientations (i.e., eye orientations with both a horizontal and vertical angular component relative to the primary position). Here we took advantage of oblique eye orientation-induced ocular torsion to examine perisaccadic orientation perception. First, we found that orientation perception was largely predicted by the rotated retinal image. Second, we observed a presaccadic remapping of orientation perception consistent with maintaining a stable (but spatially inaccurate) retinocentric perception throughout the saccade. These findings strongly suggest that our seamless perceptual stability relies on retinocentric signals that are predictively remapped in all three ocular dimensions with each saccade.

Misperception of motion in depth originates from an incomplete transformation of retinal signals

Depth perception requires the use of an internal model of the eye-head geometry to infer distance from binocular retinal images and extraretinal 3D eye-head information, particularly ocular vergence. Similarly, for motion in depth perception, gaze angle is required to correctly interpret the spatial direction of motion from retinal images; however, it is unknown whether the brain can make adequate use of extraretinal version and vergence information to correctly transform binocular retinal motion into 3D spatial coordinates. Here we tested this hypothesis by asking participants to reconstruct the spatial trajectory of an isolated disparity stimulus moving in depth either peri-foveally or peripherally while participants' gaze was oriented at different vergence and version angles. We found large systematic errors in the perceived motion trajectory that reflected an intermediate reference frame between a purely retinal interpretation of binocular retinal motion (not accounting for veridical vergence and version) and the spatially correct motion. We quantify these errors with a 3D reference frame model accounting for target, eye, and head position upon motion percept encoding. This model could capture the behavior well, revealing that participants tended to underestimate their version by up to 17%, overestimate their vergence by up to 22%, and underestimate the overall change in retinal disparity by up to 64%, and that the use of extraretinal information depended on retinal eccentricity. Since such large perceptual errors are not observed in everyday viewing, we suggest that both monocular retinal cues and binocular extraretinal signals are required for accurate real-world motion in depth perception.

Neck muscle spindle noise biases reaches in a multisensory integration task

Reference frame transformations (RFTs) are crucial components of sensorimotor transformations in the brain. Stochasticity in RFTs has been suggested to add noise to the transformed signal due to variability in transformation parameter estimates (e.g., angle) as well as the stochastic nature of computations in spiking networks of neurons. Here, we varied the RFT angle together with the associated variability and evaluated the behavioral impact in a reaching task that required variability-dependent visual-proprioceptive multisensory integration. Crucially, reaches were performed with the head either straight or rolled 30° to either shoulder, and we also applied neck loads of 0 or 1.8 kg (left or right) in a 3 × 3 design, resulting in different combinations of estimated head roll angle magnitude and variance required in RFTs. A novel three-dimensional stochastic model of multisensory integration across reference frames was fitted to the data and captured our main behavioral findings: 1) neck load biased head angle estimation across all head roll orientations, resulting in systematic shifts in reach errors; 2) increased neck muscle tone led to increased reach variability due to signal-dependent noise; and 3) both head roll and neck load created larger angular errors in reaches to visual targets away from the body compared with reaches toward the body. These results show that noise in muscle spindles and stochasticity in general have a tangible effect on RFTs underlying reach planning. Since RFTs are omnipresent in the brain, our results could have implications for processes as diverse as motor control, decision making, posture/balance control, and perception. NEW & NOTEWORTHY We show that increasing neck muscle tone systematically biases reach movements. A novel three-dimensional multisensory integration across reference frames model captures the data well and provides evidence that the brain must have online knowledge of full-body geometry together with the associated variability to plan reach movements accurately.

Transition from Target to Gaze Coding in Primate Frontal Eye Field during Memory Delay and Memory-Motor Transformation

The frontal eye fields (FEFs) participate in both working memory and sensorimotor transformations for saccades, but their role in integrating these functions through time remains unclear. Here, we tracked FEF spatial codes through time using a novel analytic method applied to the classic memory-delay saccade task. Three-dimensional recordings of head-unrestrained gaze shifts were made in two monkeys trained to make gaze shifts toward briefly flashed targets after a variable delay (450-1500 ms). A preliminary analysis of visual and motor response fields in 74 FEF neurons eliminated most potential models for spatial coding at the neuron population level, as in our previous study (Sajad et al., 2015). We then focused on the spatiotemporal transition from an eye-centered target code (T; preferred in the visual response) to an eye-centered intended gaze position code (G; preferred in the movement response) during the memory delay interval. We treated neural population codes as a continuous spatiotemporal variable by dividing the space spanning T and G into intermediate T–G models and dividing the task into discrete steps through time. We found that FEF delay activity, especially in visuomovement cells, progressively transitions from T through intermediate T–G codes that approach, but do not reach, G. This was followed by a final discrete transition from these intermediate T–G delay codes to a “pure” G code in movement cells without delay activity. These results demonstrate that FEF activity undergoes a series of sensory–memory–motor transformations, including a dynamically evolving spatial memory signal and an imperfect memory-to-motor transformation.