Amblyopia is, aside from refractive error, the largest cause of vision loss in infants and young children. It also provides a window into the effects of atypical visual development. This article traces my decades long interest in trying to understand amblyopia from the perspective of a psychophysicist, outlining some of the key questions and findings and the factors that made my journey possible. It's been quite a journey!
This study investigated whether saccadic search exists on a micro scale. Using a high-speed, confocal retinal eye tracker (FET), participants searched for tiny targets within dense displays subtending less than 4 deg2. The participants exhibited goal-directed, exploratory scanning behavior with tiny saccades (median amplitude < 0.5°); saccade amplitude scaled with inter-element distance, fixation duration increased in denser displays, and consecutive saccades followed typical directional dependencies (saccadic momentum and facilitation of return). Search performance and scanpaths resembled those found in large-scale visual search tasks, suggesting functional equivalence across spatial scales. These findings support the view that microsaccades serve the same perceptual and attentional roles as larger saccades. We argue that the distinction between microsaccades and saccades is unnecessary when describing natural, task-driven visual scanning behavior.
Most contemporary accounts of selective attention assume limited perceptual resources that need to be allocated through attention. These accounts, however, differ in how exactly irrelevant stimuli are filtered out. Suggestions include passive exhaustion of resources (e.g., load theory) or actively shaping the attended region (attentional window) such that irrelevant items remain excluded (e.g., attentional zoom account). We use pupillometry to test predictions of these accounts: Observers are asked to simultaneously monitor and compare two streams of rapidly presented items, either in a low-load (compare color) or high-load (compare the conjunction of color and orientation) condition. Flickering areas are placed between or beside the streams, falling either inside or outside the postulated attentional window. Observers respond faster and more accurately in the low-load condition, with no behavioral effects of the flicker position. In contrast, pupil size is more strongly modulated by the flicker in the inside than in the outside condition, but the modulation strength is independent of load. Exploiting the fact that the strength of the pupil modulation at the flicker frequency is an index of covert attention to the flickering region, we interpret this result as evidence for a continuous attentional window that encompasses the space between the streams. Together, our data provide evidence for an active filter mechanism, supporting theories like the attentional zoom account, rather than a passive, load-dependent spreading of residual attentional resources.
Impaired binocular vision early in life from amblyopia or strabismus is related to lower scores than controls for walking on standardized tests of motor ability. However, standardized tests do not assess gait kinematics, which may provide a better understanding of motor deficits. This study investigated gait kinematics while walking in children with amblyopia or strabismus. Children ages 7-13 years with amblyogenic factors (13 anisometropia, 8 strabismus) were enrolled in the amblyopia or strabismus (AS) group (15/21 had current amblyopia). An age-similar group of 33 controls with normal vision were also enrolled. While viewing binocularly, children walked on a GAITRite® pressure-sensitive walkway and completed 3 conditions: 1) Straight Walk (SW): walk on mat, 2) Isolated Target Walk (IT): walk and step on two-dimensional targets, and 3) Distractor Target Walk (DT): walk and step on two-dimensional targets while avoiding two-dimensional distractors. Gait kinematics were mean normalized velocity, cadence, step time, stance time, step length, step width, and step accuracy (# targets missed, # distractors stepped on). Variability in gait kinematics was also examined using the coefficient of variation (COV). Typical mean gait kinematics did not differ between the AS and control groups. However, the AS group missed more targets, showed increased variability (higher COV), and lacked any learning effects compared to controls. Unbalanced visual input early in life from amblyopia or strabismus results in variable and inaccurate walking patterns, pointing to an immaturity of gait development.
Hering's Law of binocular eye movement control guides most oculomotor research and supports diagnosis and treatment of clinical eye misalignment (strabismus). It states that all eye movements are controlled by a unitary conjugate signal and a unitary vergence signal that sum. Recent evidence of temporally asynchronous eye rotations during vergence (Chandna et al., 2021) challenges the viability of a unitary vergence signal. Helmholtz proposed an explanation of binocular control that does not require a vergence system, that the eyes are controlled independently. Yet independent control fails to explain “Remarkable Saccades”, inappropriate saccades that occur from an eye aligned on a target during asymmetric vergence (Enright, 1992). Here we present a novel computational architecture of binocular eye movement control based on a concept proposed in Chandna et al. (2021). Our “Hybrid Binocular Control” (HBC) model incorporates simplified pursuit and saccadic components. The pursuit component is implemented with independent controllers for each eye that interact with the saccade component, a unitary conjugate controller. The model generates remarkable saccades as an emergent property, and a variable attentional gain allows it to generate behavioral variations in remarkable saccades that occur when observers attend to one eye's view. In its current form, the model's output matches not only the remarkable saccade profiles well, but other saccade types. However, it does not attempt to specify their exact dynamics or underlying circuitry. The model's novel architecture operates without a vergence signal and makes predictions about how conjugate and independent controller signals interact. Furthermore, it suggests exciting exploration of neural oculomotor circuitry and inspires systematic investigations of asymmetric eye movements and their resulting main sequences necessary for further model elaboration.
Human color perception is trichromatic, and color spaces that quantify the range of perceived colors are commonly mapped in three coordinate dimensions. The human visual system can abstract a single color representation from spatial color variation, but such representations do not always conform to a colorimetric average. Here, we investigate how spatial variation in lightness, chroma, and hue, as well as their mutual relationships, affect perceptual color averages. To obtain baselines, we estimated simple discrimination thresholds for lightness, chroma, and hue. Next, we used simple, spatial color ensembles with uncorrelated and correlated variation in lightness, chroma, and hue to investigate their interactions. First, in a comparison experiment, polychromatic stimuli were compared with uniformly colored stimuli. Second, in an adjustment experiment, uniformly colored stimuli were matched to polychromatic stimuli. Biases in the perceptual averages depended on the hue range of color distributions, dimensions of color variation, as well as their intercorrelation. We found a consistent bias towards higher-than-average chroma in line with previous reports. In contrast, we found little evidence for a bias towards higher-than-average lightness previously reported for real objects. Average percepts of hue were sometimes biased towards hues correlated with higher chroma and/or lightness, and these effects varied between stimulus colors and experiments. This inconsistency in hue biases suggests that they are not a byproduct of selective subsampling or weighting of the stimulus elements with higher chroma or lightness. Even with simple, abstract stimuli, ensemble percepts of color are complicated by interactions between different dimensions of color variation.
Autistic people are less likely to obtain a driver's license than neurotypical peers and may face higher collision risks when driving. While advanced driver assistance systems (e.g. hazard warnings) could enhance driving safety for this population, their efficacy remains understudied. In this study, licensed drivers with low and high levels of autistic traits, and non-autistic controls, completed simulated city drives, avoiding pedestrian hazards with and without a vibrotactile warning device. Gaze fixation, braking responses, and subjective workload were compared across warning conditions and between groups. Without warnings, autistic trait level had no effect on gaze fixation or brake responses to hazards. However, when driving with warnings, drivers low in autistic traits exhibited delayed initial fixation, shorter total fixation duration, and slower braking responses, but did not have increased collisions. While drivers with high autistic traits showed no behavioral changes when using warnings and reported high frustration, they achieved the greatest reduction in collisions. Unique gaze patterns were seen among drivers high in autistic traits, which might suggest differing adaptive behaviors developed by individuals in this group to manage hazard situations. These findings highlight the potential of driver assistance systems to enhance safety in this population and demonstrate the need to consider autistic trait levels when evaluating driving and the potential benefits and limitations of such driver assistance systems.
Attention-Deficit/Hyperactivity Disorder (ADHD) is typically conceptualized as a disorder of executive control; however, accumulating evidence suggests that early sensory processing may also be atypical. The present study examined contrast sensitivity (CS) and noise processing in a sample of 45 adults (21 with ADHD and 24 neurotypical controls). In Experiment 1, foveal CS was measured across spatial frequencies (4-12 cpd) using standard-size targets presented at stimulus durations of 40 and 80 ms. A three-way ANOVA revealed a significant main effect of group, indicating overall lower contrast sensitivity in the ADHD group, but no significant interactions between Group, Spatial Frequency, and Stimulus Duration. To examine the effect of stimulus size, an enlarged stimulus condition was introduced at the highest spatial frequency (12 cpd). At the 40 ms duration, a significant Group × Stimulus Size interaction suggested reduced spatial integration in the ADHD group, with controls showing greater improvement when stimulus size increased. At 80 ms, this interaction was not significant, although the main effect of group remained present. In Experiment 2, perceptual noise exclusion was assessed; overall performance did not significantly differ between groups across noise levels and stimulus durations. Together, these findings suggest that visual differences in ADHD are selective, emerging primarily under conditions requiring efficient integration of high spatial frequency information at short stimulus durations, while robust group differences in external noise exclusion were not detected.
We show a 90° shift in the perceived direction of a well-known motion illusion, based on the dot-pair separation in dynamic Glass patterns. The change can be explained with a cortical model incorporating extra-classical receptive fields. We employed dynamic Glass patterns: a sequence of patterns of randomly positioned, but consistently oriented dot-pairs, presented in rapid succession. Although dynamic Glass patterns contain no coherent motion, observers perceive strong motion in either direction parallel to the dot-pair orientation for small dot-pair separations (< 30 arc min). This has been attributed to the visual system interpreting the dot-pairs as motion streaks of fast-moving objects. The motion illusion has been explained by a model of the orientation selectivity of simple cells in primary visual cortex for small dot-pair separations. However, those models did not include the influence of end-stopping, an extra-classical receptive field mechanism that enhances length-selectivity to oriented bars. We incorporated end-stopping into the model and showed that increasing dot separation in Glass patterns shifts the directional preference of simple cells from parallel with the dot-pairs to the orthogonal direction. Our psychophysical experiments confirmed that Glass patterns perceived as rotating, with small dot-pair separations, were perceived as expanding/contracting with large separations (> 53 arc min). Furthermore, this shift to radial motion was eliminated when dot-pairs with opposite contrast polarities were used, consistent with our end-stopping model. The results provide new insight into the interaction between motion and form cues in the visual system, highlighting the role of extra-classical receptive fields in motion perception.
In the peripheral drift illusion, a static pattern containing asymmetrical spatial luminance gradients can appear to move at its onset and then whenever the viewer moves their eyes or blinks. Recent evidence indicates that the illusion is due to changes in retinal luminance created by pupillary reflexes. Can computational models of human cortical motion sensing account for the illusory motion? An implementation of the Adelson-Bergen motion-energy models was unable to account for the illusion. However, a variant of the model incorporating half-wave rectification (half-squaring) applied to the response of the model's direction-selective sensors can account for the illusion. Half-squaring creates parallel ON and OFF channels. The output of the OFF channel was found to signal motion in the direction of the illusion whereas the ON channel signalled the opposite direction. If the two channels' outputs were combined in a way that is biased in favour of the OFF channel, the net output accounted for the illusion. A bias in favour of OFF responses is supported by evidence from physiological and psychophysical studies of human vision.
Over the last two decades, models of visual processing in the retina have increased in scale and predictive power, driven by advances in recording technologies and deep learning. Current models span a spectrum from functional, high-performance architectures to detailed biophysical simulations. Here, we survey the state-of-the-art across this spectrum, with a focus on the key developments needed to leverage these models to improve our understanding of retinal processing. We argue for a collective effort to build a scientific ecosystem around a core of shared datasets, modular composable models, and standardised benchmarks. We propose that such an infrastructure would accelerate model-driven discovery and enable the community to focus its efforts on open scientific questions such as the role of neural variability, adaptation, and cell-type diversity in retinal coding.
This study investigated foveal crowding and segmentation effects in adults with Attention- Deficit/Hyperactivity Disorder (ADHD) compared to neurotypical controls. While visual crowding, the impaired identification of targets surrounded by flanking elements, has been extensively studied in peripheral vision across various populations, its manifestation in central (foveal) vision among adults with ADHD remains insufficiently studied. 45 participants (22 ADHD, 23 controls) completed a computerized visual task involving identification of the direction of a central target letter ("E") under crowded and uncrowded conditions. A subset of 26 participants (14 ADHD, 12 controls) completed an additional segmentation condition in which the target appeared in red while flankers remained black. Accuracy and response times were recorded across varying stimulus durations (30-240 ms). Results indicated that adults with ADHD exhibited significantly stronger foveal crowding Effects and slower response times than controls. Segmentation improved performance in both groups. These findings suggest that perceptual alterations associated with ADHD extend to basic perceptual processes in central vision. The enhanced crowding reflects a reduced ability to filter irrelevant visual information and an increased integration field, which heightens interference between target and flanker signals and thus increases crowding susceptibility. Salient visual cues mitigated these effects by enhancing target distinctiveness. The findings have implications for everyday visual behaviors, including reading and driving in complex environments, where impaired filtering of distracting information may affect performance.
A2 amacrine cells (AIIACs) are the most important amacrine cells (ACs) in the mammalian retina, as they are the "hubs" for merging rod and cone signals in the inner retina and segregate them into ON and OFF ganglion cell (GC) pathways. It is unclear, however, whether AIIACs send their signals to all types of GCs in the mammalian retina. A major reason for this knowledge gap is that it has not been possible to selectively stimulate AIIACs without stimulating other retinal neurons. In this study, we will use a newly-developed mouse retinal preparation in which channelrhodopsin (a ChR2 [H134R]-Venus fusion protein) is selectively expressed in AIIACs via intravitreal injection of AAV serotype 2 (AAV2) carrying the ACAGW-ChR2-Venus construct (Addgene 20,071). We have identified a titer range within which ChR2 expresses ONLY in AIIACs and not in any other cells, yet large and clear ChR2-elicited postsynaptic signals can be observed. We examined effects of ChR2-elicited AIIAC depolarization on four types of alpha GCs, the ON-sustained αRGC (sONαGC), OFF-sustained αRGC (sOFFαGCs), ON-transient αRGC (tONαGCs) and OFF-transient αRGC (tOFFαGCs), and found that AIIAC depolarization mediates the sONαGC and sOFFαGC responses via the AIIAC-cone DBCR/MC (rod/M-cone ON Bipolar Cell) gap junction pathway and the AIIAC-OFFαGC glycinergic pathway, respectively. AIIAC depolarization did not affect the tONαGCs and tOFFαGCs, whose responses might be mediated by other ACs. These experiments demonstrate that AIIAC depolarization mediates the responses of a subpopulation of ON and OFF GC types, suggesting that signals from any single retinal neuron type participate in restricted, cell-type-specific circuitries linked to retinal output pathways.
Perceptual learning, particularly with texture discrimination training (TDT), offers a valuable framework for investigating sensory and cognitive plasticity in clinical populations. In this study, we examined the effects of TDT in adults with attention deficit hyperactivity disorder (ADHD), with and without pharmacological medication treatment. Participants performed a dual-task paradigm involving central fixation identification and peripheral texture orientation discrimination, across four training sessions. Participants with ADHD without medication exhibited elevated perceptual thresholds and slower reaction times on the texture discrimination task, relative to neurotypical controls. In contrast, ADHD participants with medication demonstrated significantly lower thresholds, even surpassing those of the control group. Despite these baseline differences, all groups showed perceptual learning, with improved performance over the days. These findings suggest that while overall perceptual learning mechanisms in ADHD are comparable to those of controls, there are distinct differences in temporal processing and attentional allocation, particularly under dual-task conditions.
Individuals vary considerably in their ability to recognize faces. Such variability may in part reflect individual differences in exposure to faces based on the number or diversity of faces that are available to be seen in the environment. To investigate this possibility, a series of studies have used hometown size as a proxy for individual variation in face exposure and investigated whether it predicts individual differences in face recognition ability, with mixed results. In particular, it was previously found that performance on face recognition tasks was predicted by hometown size, but only for male participants. In three experiments, we sought to revisit whether hometown size predicts face recognition ability, using an order of magnitude larger sample sizes (from N = ∼90 increased to N = ∼ 900), while also considering a variety of other predictors related to childhood experience (e.g. neighborhood type and family composition) and adult sociality, including personality, social intelligence, and shyness. Across all experiments, we found that hometown size was a poor predictor of adult face recognition ability compared to other predictors related to childhood experience. In addition, of the different measures related to sociality, social awareness was a comparatively strong predictor of individual differences in face recognition ability.
Peripheral vision plays a crucial role in visual recognition, with previous studies suggesting that cognitive predictions derived from peripheral input enhance object categorization in central vision during single eye-fixation. However, visual perception is dynamic, alternating between ocular fixations and saccades to new objects of interest. In three eye-tracking experiments, we explored how task-relevant predictions extracted from the peripheral scene context influence the gaze-based discrimination of target objects in peripheral vision. We presented both an object and a scene in peripheral vision and instructed participants to saccade toward the object if it belonged to a predefined target category, or to maintain central fixation if it did not. We manipulated the semantic congruence between the peripheral scene and the target category, and we measured its effect on saccades. Across the three experiments, the results consistently showed that a scene congruent with the target category triggered erroneous saccades toward the distractor object. Experiment 2 revealed that the congruence effect was amplified when the scene was presented before the object, suggesting that the accumulation of information about the peripheral scene contributed to generating predictions and thus increased the likelihood of saccades toward the peripheral object. Experiment 3 further clarified that scene context alone did not account for the erroneous saccades observed in previous experiments, and that these responses reflect the combined influence of scene context and object presence. This study demonstrated that the peripheral scene context, although irrelevant to the task, was automatically processed and used to form predictions that trigger saccades toward expected objects in peripheral vision.
Symmetry, a fundamental concept in nature, science and art, has challenged computer vision researchers because it occurs in various forms and human symmetry perception can deviate from the mathematical definition. Previous symmetry detection datasets are limited by the number of annotators and by missing the nuances of human perception. We introduce PIX2PER, a novel dataset for reflection symmetry in natural scenes and artworks. We also introduce WF1, a modified version of the widely-used F1 detection performance score, by adding weights to precision and recall to accommodate for the perceived symmetry strength. Created by adding weights to precision and recall to accommodate for the perceived symmetry strength. We perform a comparative analysis of existing models for symmetry detection on this human-centric dataset. Additionally, we present a fully synthetic dataset for pretraining symmetry detection models. When finetuning this pretrained model with human data, performance increases significantly. This research introduces and evaluates ways of improving symmetry detection and contributes to the development of computer vision models that more effectively represent human perception.
This study investigated the relationship between form-deprivation myopia and amblyopic deficit. Five rhesus monkeys underwent monocular form-deprivation beginning at 24 ± 2 days of age. Lightweight helmets held a light-perception only Bangerter filter over the right eye and a plano lens over the left eye. Refractive error and axial length were measured biweekly using streak retinoscopy and noncontact biometry. After 135 ± 2 days of treatment, visual function in each eye was assessed under sedation by recording Visual Evoked Potentials (VEP) elicited by monocular stimulation with square wave gratings (spatial frequencies 0.25-16 cycles per degree). An Ocular Dominance Index (ODI), the difference in VEP amplitudes between form-deprived (amblyopic) eye stimulation and fellow (control) eye stimulation divided by the sum, quantified amblyopic deficits. At the end of treatment, mean spherical equivalent refraction and axial length were -0.43 ± 4.41 D and 16.66 ± 0.70 mm in form-deprived eyes and +1.75 ± 2.71 D and 16.07 ± 0.61 mm in fellow eyes, respectively. Anisometropia was highly variable, ranging from -6.13 to +2.75 D (mean -2.18 ± 3.60 D). All monkeys demonstrated a significant amblyopic deficit, with a mean ODI of -0.66 ± 0.11 (range: -0.43 to -0.81). Correlations between ODI and anisometropia and ODI and axial length were not significant in this small cohort. Monocular form-deprivation in young rhesus monkeys produced a robust amblyopic deficit as measured by VEPs, even in the absence of myopia. Findings indicate that form-deprivation induces central visual pathway deficits may not be correlated with the amplitude of form-deprivation myopia.
