Abstract
Keynote: Investigating the human binocular visual system using multi-modal magnetic resonance imaging
University of Oxford, UK
Email:
Seeing in stereoscopic depth relies on the slight differences in the images in the two eyes resulting from their horizonal separation. As the inputs from the eyes are first combined in the primary visual cortex (V1), human brain imaging can be used to investigate the earliest stages of processing. To perceive depth, binocular neurons in V1 need balanced input from the two eyes, likely established by mutual inhibition, whereby activation of one eye inhibits the other eye. Thus, when input from the two eyes is comparable, both will contribute equally to binocular vision to promote 3D vision. However, if input from one eye is consistently weaker, the other eye may dominate, a situation that can lead to amblyopia. I will describe recent work using magnetic resonance spectroscopy indicating a role for inhibitory neurotransmitter GABA in binocular vision and eye dominance. I will then present two studies that investigate how population receptive field (pRF) mapping can determine the selectivity of cortical areas to binocular disparity. In the first study we show that V1 pRFs measured with a disparity-defined stimulus are larger than those generated by stimuli that do not require binocular interaction. Secondly, using pRF mapping in the depth plane, we show a relationship between the preferred depth and width of the tuning curve, such that voxels preferring zero disparity have narrower tuning curves. The ability to understand the mechanisms underlying human binocular vision raises the possibility of manipulating neural activity levels to improve function in those with impaired binocular vision.
The impact of auditory, visual, and audio-visual sensory cues on multiple object tracking in children.
1University of Lincoln, UK
2University of Staffordshire, UK
Email:
Previous studies have shown that the provision of additional auditory and visual cues allow adults to track a higher number of moving objects amongst similar distractor objects compared to no-cue conditions (Föcker et al., 2022). Here, we investigated the developmental trajectory of the ability to integrate sensory cues during multiple object tracking in 6–10 year old children. Interestingly, multisensory integration abilities in children optimise at around 8–10 years and children younger than 8 years show task-specific uni-sensory dominance (Gori et al., 2008). In two experiments, children were asked to track multiple moving target objects that elicited auditory, visual, audio-visual cues when they bounced against an inner circle. In Experiment 1, four target objects among four distractor objects were presented. Results revealed that children showed higher tracking performance when visual cues were present compared to when visual cues were absent. This ability to use visual cues to enhance tracking increased with age. To further explore the integration of auditory cues during tracking, in Experiment 2, cognitive load was reduced by presenting two target objects among six distractor objects and an auditory cue was either present on collision with the inner orange boundary or absent. Results revealed that auditory cues did not improve tracking performance irrespective of the load condition (four versus two targets). We discuss possible underlying mechanisms of these findings such as cross-sensory calibration, cognitive load and multisensory integration abilities in children.
Similar dynamics of normalization reweighting for monocular and dichoptic masks
University of York, UK
Email:
Recent neurophysiological and modelling work has overturned three decades of orthodoxy by showing that suppression between populations of neurons can be ‘reweighted’ by recent history (Aschner et al., 2018 Journal of Neuroscience 38, 10129–10142). Here we investigate the temporal dynamics of this normalization reweighting using steady-state EEG and MEG. By tagging target and mask components at different frequencies (5 and 7 Hz), we measure the timecourse of suppression between orthogonal component gratings. Across three datasets, with a combined sample size of N = 220 adults, we find that: (i) suppression increases during the first 2–5 s of stimulus presentation, and (ii) the dynamics are similar for monocular and dichoptic presentation. The existence of dichoptic reweighting effects allows us to rule out a major contribution from early (pre-cortical) stages of processing. Using MEG source-localization, we confirm that these effects are present in V1, consistent with neurophysiological results. Given longstanding theories about differences in inhibition associated with autism, we also investigated whether normalization reweighting varied with autism diagnosis, or with autistic traits in the general population. However this analysis revealed no compelling differences. Overall, our findings support the normalization reweighting model, reveal the timecourse of the effect for the first time, and demonstrate that reweighting also occurs for dichoptic masks.
Modulation of visual acuity induced by contrast reversal of targets and backgrounds
University of Nottingham, UK
Email:
It has recently been shown that static and dynamic acuity in the peripheral visual field can be improved by reversing the contrast polarity of the target (Patrick, Roach & McGraw, 2019 Journal of Vision, 19(13):12, 1–19). However, it is not clear whether this advantage is specific to modulation of the target, or whether manipulations of the background might yield similar improvements. Here we tested peripheral acuity (10 degrees eccentricity) for modulated (10 or 15 Hz) and unmodulated Landolt C targets presented on either a uniform background, static patterned background (1/f2 noise with 90% contrast) or a patterned, contrast-reversing background (10 or 15 Hz). Nine individuals with normal vision and one individual with central vision loss were tested. Observers were required to judge the orientation of the target (four alternatives). To estimate thresholds, we manipulated target size using a staircase procedure and fitted full psychometric functions to data aggregated over several blocks. Results indicated that stimulus configurations containing dynamic change frequently led to better performance than static displays. However, while modulating the target yielded optimal performance for some observers, modulating the background was more effective for others. These findings indicate that introducing temporal modulation may hold promise for improving visual function of patients with central vision loss, who are forced to rely on peripheral regions for spatially demanding visual tasks. However, further work will be required to optimise the form of temporal modulation for a given individual and task.
The impact of high-contrast linear floor patterns on human gait kinematics
University of Bristol, UK
Email:
Visual patterns which deviate from the statistical properties of natural scenes result in inefficient processing in the visual cortex, causing visual discomfort. In particular, high-contrast, square-wave grating patterns produce the strongest perceived visual discomfort. This effect is most pronounced for gratings around three cycles-per-degree of visual angle (CPD). Many patterns fitting this description can be seen in the urban environment. However, to date, there is no investigation which examines whether such patterns, when present on the floor, affect fundamental characteristics of human gait. Here, participants (n = 60) were asked to walk over high-contrast, linear patterns projected onto the floor, and to subjectively rate each pattern for visual discomfort. The CPD was parametrically varied (1.0, 0.5, 0.25, 0.125, 0.0625, 0.03125 CPD) trial-by-trial in random order across three pattern configurations: orthogonal, parallel and oblique to walking direction. ANOVAs and multilevel modelling demonstrated that velocity and step length decreased as spatial frequency increased. Patterns with 1.0 and 0.5 CPD led to the most prominent reduction in velocity and step length in comparison to 0.0625 CPD patterns and a medium-luminance, grey control floor. However, pattern orientation did not appear to significantly influence changes in gait. Subjective ratings of visual discomfort were negatively correlated with changes in velocity and step length. We discuss the impact of high-contrast linear floor patterns with regard to visual discomfort, perceptual load and the consequences this has for the fundamental characteristics of human gait. Moreover, we consider the implications this might have for accessibility and future urban design.
Eye tracking the effect of turbulence in particle systems in virtual reality
Royal Holloway University, UK
Email:
A 3D immersive experience allows the user to choose how they sample their environment, however, often the designer wishes the user to view certain parts as part of the experience. A particle system is a technique in game engines which can be used in virtual reality for the subtle formation of virtual objects composed from small dynamic components. In this work, eye tracking was used to investigate the effect of this simulation technique as a method of visual attention guidance within a 360° immersive experience. We made use of a custom created scene from a narrative piece as a ‘real world’ case study. We compared the effect of what is termed ‘turbulence’ – a setting for particles that manipulates the amount of noise (jittering away from a pre-set path) in their 3D trajectories. We compared two different settings for a ‘swarm’ particle cloud that formed part of a scene containing a cityscape and a character generated by 3D scanning of an actor addressing the viewer. We found that there was a small but significant advantage for the less turbulent particle cloud to hold attention and in general the swarm was attended to 6% more of the time when comparing total fixation duration over the whole movie duration. This study shows how eye tracking in 3D can be used to choose between different properties of the scene for increasing the effectiveness of attention guidance and further exploratory data indicated that this difference in effectiveness varied with what was visible to the viewer.
Chromatic discrimination measures in mature observers improve with a longer response window
1University of Oxford, UK
2Federal University of Pará, Brazil
3University Hospital Erlangen, Germany
4Liverpool Hope University, UK
Email:
In the present study, the Trivector test (Cambridge Colour Test) was employed to assess chromatic discrimination in normal trichromats (N = 30) aged 20–64 years, stratified as ‘young’ (20–29), ‘middle-aged’ (31–48) and ‘mature’ (57–64). Earlier our anecdotal evidence prompted that a longer response window (RW) improved older observers’ estimates. Here, we systematically probed the RW duration – 3, 5 and 8 s, to explore whether its variation would affect chromatic discrimination in Protan, Deutan and Tritan subsystems. For the ‘young’ and ‘middle-aged’ subgroups, we found no differences for measures between all RWs. However, the RW effect was apparent for the ‘mature’ observers: at 3-s RW, all measures were significantly higher than at 8-s RW. Notably, at 3-s RW the Tritan measures were greatly beyond the upper tolerance limit for this age group (Paramei & Oakley, 2014 JOSA A 31 A375-A384), spuriously indicating impaired chromatic discrimination. Furthermore, Protan and Tritan measures decreased at 8-s compared to 5-s RW, while Deutan measures levelled at 5-s RW. The elevation of chromatic discrimination thresholds at shorter RWs is indicative of an increase in critical processing duration with age. Crucially, the findings imply that for accurate performance, older observers require longer stimulus exposure, to ensure veridical measures of chromatic discrimination by disentangling these from ageing effects – slowing down of visual processing and motor responses (Fars et al. 2022 Sci Rep 2(1): 9072). Based on our findings, we advocate a sensible RW extension in the Trivector protocol for testing mature observers.
The effects of moderate and profound visual deprivation on auditory emotional distractor processing in a serial recall task
1University of Lincoln, UK
2Technische Universität Darmstadt, Germany
3University of Geneva, Switzerland
4Health and Medical University, Germany
Email:
Previous research suggests that blind individuals are better able than sighted individuals to detect target syllables while ignoring distractor syllables when emotional prosody is task irrelevant (Topalidis et al., 2019). Furthermore, blind individuals’ short-term memory has been reported to be less susceptible to distraction from irrelevant speech compared to sighted individuals, indicating their improved ability to suppress the encoding of irrelevant sounds (Kattner & Ellermeier, 2014). Here, we tested whether the emotional prosody of task-irrelevant auditory syllables modulates distraction by irrelevant sounds and whether the degree of visual deprivation (i.e. moderate versus profound) predicts resistance to distraction. Seventeen sighted participants, 18 visually impaired individuals and 17 blind individuals were asked to memorize spoken digits in serial order either during a period of silence or while pseudo-words were spoken with different prosodic intonations (happy, neutral, fearful, threatening; steady-state sounds). Results suggest that blind listeners’ short-term memory capacity is higher compared to visually impaired and sighted individuals, regardless of the irrelevant sound condition, whereas sighted and visually impaired individuals did not differ in performance. Blind individuals’ recall accuracy did not vary as a function of the emotional prosody of steady-state sound, whereas visually impaired individuals showed higher performance in the threatening condition compared to all other emotions. In contrast, sighted individuals’ serial recall accuracy was higher in the fearful prosody compared to neutral and threatening prosody of irrelevant speech. Enhanced top-down control and reduced attentional capture by emotional prosody are discussed as explanatory mechanisms of enhanced performance in blind individuals.
The effects of segmenting spatial structure on time-limited foraging tasks
Nottingham Trent University, UK
Email:
Visual foraging can involve searching a display for multiple instances of targets. The spatial structure of a display affects search behaviours in classic search tasks (Nakashima & Yokosawa, 2013, AP&P, 75, 299–307). We recently found that spatial structure has also recently been shown to affect foraging. It is unclear whether this effect is unique to exhaustive foraging, and therefore this was investigated in two experiments. Participants foraged segmented and non-segmented displays for as many targets as they could in 20 s without being required to forage exhaustively. Displays contained 40 distractors and either 40 feature targets or 40 conjunction targets. Targets disappeared after selection in Experiment 1. Experiment 2 was more difficult as targets remained visible after selection. In both experiments, participants organised their foraging differently and incurred costs when foraging segmented displays, relative to non-segmented displays. That is, individuals organised their foraging in a segment-by-segment manner in Experiments 1 and 2. In Experiment 2 participants made fewer erroneous re-clicks on targets in segmented displays compared to non-segmented displays. Conjunction foraging was associated with longer foraging paths and fewer target-type switches than feature foraging, replicating previous results (Kristjansson et al., 2014, Plos One, 9, e100752). Individuals might forage displays in a segment-by-segment manner because this structure might facilitate systematic search activities in non-exhaustive foraging, just as it did in exhaustive foraging. Individuals may also find it effortful to switching attention between display segments, and so may forage in a segment-by-segment to avoid these costs.
Narrative, not low-level vision, synchronises audiences during television viewing
1University of Bristol, UK
2Bristol Vision Institute, University of Bristol, UK
3BBC Research and Development, UK
Email:
Cinematic media (e.g. film, television) possesses a remarkable ability to synchronise audiences’ neural, behavioural, and physiological responses. This synchrony, sometimes termed ‘tyranny of film’, is largely considered to arise from low-level visual features and editing conventions. Recently, evidence has suggested that synchrony may also emerge from shared interpretation of narrative. However, no study to date has assessed the relative contributions of narrative and low-level features towards synchrony. We designed a study in which participants (n = 60) were presented with a 55-min episode of the BBC television drama The Tourist. Content was presented to participants in one of two modalities: audio-only with audio-description, or visual-only with subtitles. In this way, the presentations shared no low-level features, but participants experienced the same narrative. During the sessions, we recorded participant's heart rate, and computed synchrony from this physiological measure using intersubject correlation analysis. We found evidence that synchrony was higher in the audio versus visual condition, however no significant differences were found between either visual or audio conditions and synchrony between-groups (i.e. narrative condition). Further, when modelling heart rate, 22% of variance could be explained by narrative, compared to 1.7% from low-level saliency. Saliency was derived through Itti, Koch, & Niebur's (1998) saliency model, and computing root mean square energy from the audio track. Our results provide strong support for the idea that processing of a narrative can lead to markedly similar physiological responses across an audience. This effect is likely high-level, and cannot be explained by visual or auditory salience alone.
The role of non-physical disparities in understanding binocular depth perception
University of Essex, UK
Email:
Binocular information is encoded in the visual cortex by disparity tuned cells. Computational models play a central role in our understanding of this process, and stimuli with non-physical disparities, that are not typically encountered in the natural environment, have been important in the development of these models. An example of these stimuli is the anti-correlated random dot stereogram, in which the contrast polarity of elements in one eye is reversed. Anticorrelated stimuli have been used to evaluate and refine models of the encoding of binocular information by neurons in the primary visual cortex. In the current analysis, here we assess the role of anticorrelated stimuli in the development of models of the perception of depth. In order to predict the direction of depth perceived in anti-correlated stimuli, we need to specify the distribution of disparity tuning in populations of binocular neurons; the process via which this population response is used to estimate depth; and the roles of second-order channels and top-down feedback. We show how the predictions from these models vary with the tuning parameters of feedforward and feedback signals, and how this can be used to test theories of binocular depth perception.
Three problems and solutions for running psychophysics experiments online
1Open Science Tools Ltd, UK
2School of Psychology, The University of Nottingham, UK
Email:
The COVID-19 pandemic saw a sharp and understandable rise in the number of cognitive scientists taking their research online. Researchers from a range of fields took their tools to the internet, to continue the study of behaviour. Psychophysicists, however, faced unique challenges and problems. Here we’ll discuss three problems, and some solutions, for running psychophysics experiments online. First, visual psychophysicists often need to know the size of stimuli as they fall on the retina, in degrees of visual angle; this requires knowledge of screen resolution, screen size and viewing distance. Second, in particular for tasks of temporal perception, a high degree of control over stimulus and response timing is required in many cases. Third and finally, many stimuli in psychophysics require luminance modulation across a linear scale (e.g. gratings). This presents two problems for online studies. How can we actually set gamma values, given that a browser doesn’t have access to the hardware lookup table (LUT) and how do we measure the gamma value we require if our remote participant doesn’t have a photometer to hand? Here we’ll outline the range of innovative solutions scientists have been sharing with scientists over the last few years to enable online psychophysics in the face of this particular set of challenges. In particular, we present new data showing how PsychoJS solves the problems of gamma correction and rapid, precise psychophysical calibration.
Training Equestrian situation awareness and eye movement behaviours using expert commentary and visual cueing
Nottingham Trent University, UK
Email:
Previous research has shown that experienced or more expert athletes have improved situation awareness (SA) compared to their novice and less-expert counterparts. Yet, SA training is underdeveloped in sports. Previous research in other domains has shown that both visual cueing techniques and commentary training may produce improvements in SA-related elements (perception, understanding and prediction). The aim of this study was to therefore explore whether training using these techniques improves prediction ability in Equestrian Polo. Non-polo player participants were randomly assigned to a training intervention (visual cueing, expert commentary, control) and completed a pre- and post-training What Happens Next? Test within the context of Equestrian Polo. Accuracy scores, number of fixations, fixation durations and fixation dispersions were measured during the pre-training and post-training tests and compared. Results showed there were no differences in WHN test accuracy across the three training groups in both the pre-training and post-training WHN tasks. There were no differences in eye movement behaviours between the pre-training and post-training tests for each training group. There were general eye movement patterns observed where results showed there were increased number of fixations towards areas ‘Off-the-Ball’ compared to ‘On-or-Near-the-Ball’, with longer fixation durations at ‘On-or-Near-the-Ball’ locations. This result may highlight that high processing demands were incurred within our novice population who have not experienced polo previously. Overall results demonstrate that more nuanced, prolonged and/or scaffolded approaches to SA training, that considers player experience and domain-specific cognitive expertise level, are likely needed in complex tasks such as sports.
Cataract type dependence on lens thickness parameter
University of Latvia, Latvia
Email:
Introduction: Nowadays cataract type is diagnosed with subjective methods. Health care professionals are looking for a way to diagnose cataract type objectively (e.g. Lam et al., 2015 Nature Reviews Disease Primers 1 15014) Our aim was to use a lens thickness parameter and explore its effects on cataract type, anterior chamber depth and axial length. Method: We investigated 750 patients (mean age 73 ± 14) which were divided depending on their cataract diagnose: nuclear (n = 250), cortical (n = 250), posterior subcapsular (PSC) (n = 250). The lens thickness parameter (LT), anterior chamber depth (ACD) and axial length (AL) were measured with IOL Master 700 (Zeiss). Results: Lens thickness parameter was greater for PCS patients (mean 4.87 ± 0.14 mm). Cortical cataract LT was 4.50 ± 0.17 mm, nuclear 4.37 ± 0.15 mm. All the cataract group results were statistically significant compared to each other (p < 0.01). PSC patients had the widest ACD (mean 3.35 ± 0.12 mm), cortical cataract patients had the shallowest ACD (mean 3.09 ± 0.15 mm). There was no statistical significance between the nuclear and cortical ACD results (p = 0.14). Axial length was longer for PSC patients (mean 24.79 ± 2.11 mm), shortest to cortical patients (mean 22.70 ± 2.77 mm). There was a statistically significant difference between all cataract groups (p < 0.01). There was no correlation between the AL and LT parameters. Conclusion: Lens thickness parameter can be used as an additional criteria to differentiate cataract types. More patient data should be collected to develop a scale that indicates particular cataract type.
Adaption-induced distance compression integrates spatial information over time
University of Nottingham, UK
Email:
Patterns of activity on the retina signal the spatial relationship between objects in the real world. However, how these patterns are used to generate spatial perception is elusive. Adaptation to a dynamic texture reduces the perceived separation between two objects presented simultaneously at that location (Hisakata et al., 2016 Current Biology 26(14) 1911–1915), modifying the mapping between physical and perceptual space. Here we ask whether this effect also occurs when objects are presented asynchronously, to test temporal properties of this mechanism. Following presentation of an adaptor (a dynamic irregular grid of dots), observers saw two pairs of dots on either side of the fixation point and chose the pair with a larger inter-dot spatial separation. In different conditions, the temporal delay between the two dots in each pair was manipulated (0, 8.3, 16.7, 25, 100, 500 or 1500 ms). The first dot was always presented at 6 degree of visual angle from fixation point on the horizontal meridian. The second dot was presented 1 degrees away from the first dot (standard) or with a variable separation (test). Orientation of the dots in the standard pair was randomly chosen on each trial, and the test pair was rotated 180 degrees relative to the standard. We found a separation compression at the adapted region (∼15% for simultaneous and ∼10% for asynchronous presentation), even when dots were separated in time (up to 100 ms). The results suggest that adaptation alters spatial computations requiring integration of information over a limited temporal interval.
The computational principles underlying the oblique effect in human vision
Univeristy of Nottingham, UK
Email:
The ‘oblique effect’ (OE) is the well-known phenomenon that we are much better at seeing horizontal and vertical (i.e. cardinal) edges than more tilted ones, but the precise mechanisms underlying this phenomenon remain elusive. To elucidate the basis of the OE in human vision, participants (N = 9) performed a series of psychophysical tasks using oriented Gabors (1–16 cpd) to measure contrast-detection, orientation-discrimination and orientation-identification performance. Participants showed an OE for contrast-detection only at high spatial frequencies, but the OE was invariant with spatial frequency for orientation-discrimination. Furthermore, OEs for orientation-identification were markedly larger than those for orientation-discrimination. To understand the potential neural basis of these results, a physiologically inspired computational model, composed of a population of orientation-tuned filters that could be corrupted by noise, was used to simulate neural responses on each task and a ‘read-out’ algorithm used to decode the model's activity. To capture the OE for contrast-detection, the model required an overrepresentation of cardinal-tuned neurons with uniformly narrower bandwidths for units tuned to higher spatial frequencies. To predict the orientation-discrimination data, the model required that cardinal-tuned neurons had a narrower orientation bandwidth than those tuned to obliques. We reproduced the pattern of OEs for the orientation-identification task in the model by comparing the decoded orientation against a noisy template (memory representation) of the standard orientation. Overall, these results suggest that OEs in human vision may arise from a combination of factors including overrepresentation of mechanisms sensitive to cardinal orientations and/or broader tuning for oblique orientations.
PySilSub – a toolbox for silent substitution
1University of York, UK
2University of Washington, USA
3Technical University of Munich, Germany
4Max Planck Institute for Biological Cybernetics, Germany
Email:
A normal human retina contains several classes of photosensitive cell – rods for low-light vision, three types of cones for daylight vision and the intrinsically photosensitive retinal ganglion cells (ipRGCs) expressing melanopsin for controlling non-image-forming functions (e.g. pupil size, circadian rhythms). The spectral sensitivities of the photoreceptors overlap significantly, meaning most lights will stimulate all photoreceptors, but to varying degrees. The method of silent substitution (Estévez & Spekreijse, 1982, Vision Research, 22[6], 681–691) provides a principled basis for stimulating individual photoreceptor classes selectively, which is useful in research and clinical settings. The main hardware requirement for silent substitution is a spectrally calibrated light stimulation system with at least as many primaries as there are photoreceptors under consideration. Device settings that will produce lights to selectively stimulate the photoreceptor(s) of interest can be found using a variety of analytic and algorithmic approaches. Here we present PySilSub, a novel Python package for silent substitution featuring object-oriented support for individual colorimetric observer models, multi-primary stimulation devices and solving silent substitution problems with linear algebra and constrained numerical optimisation. The software is registered with the Python Package Index (pip install pysilsub) and includes example data sets from various multi-primary systems. We hope that PySilSub will further encourage the application of silent substitution in research and clinical settings.
Measuring variability in EEG waveforms: lessons from visual evoked potentials
University of Edinburgh, UK
Email:
Recently, standard error of the mean (SEM) was proposed as a measure of electroencephalographic (EEG) data quality (Luck et al., 2021 Psychophysiology 58:e13793). According to this approach, those participants who have higher variations of amplitude within their single-trial responses are more noisy/variable. However, at the core of the event-related potential (ERP) method lies an assumption that observed ERPs reflect a manifestation of underlying latent ERP components, which are assumed to be either positive (e.g. P1) or negative (N1) in amplitude. ERP components may therefore be subject to rules that govern zero-bound measures. This poses a fundamental problem for SEM as a measure of variability, as in zero-bound measures those participants with higher mean responses also have more room for variability. To evaluate this, we median-split an EEG dataset with 38 participants (Martinovic et al., 2018 Journal of Vision 18 4) into two subgroups based on the amplitude of their early sensory component P1. Higher P1 amplitude reflects more phase locking to the stimulus and thus higher signal strength. We compare SEMs and coefficients of variation (CV; standard deviation divided by the mean) between the low-P1 and high-P1 groups, finding that the high-P1 group has significantly elevated SEM throughout the trial, that is, including the baseline that precedes stimulus onset and reflects non-evoked neural oscillations. Unlike SEM, CV does not identify between-group differences in variability during baseline. In conclusion, to be suitable for quantifying data noisiness, a standardised ERP variability measure should also account for the mean.
The trouble with classification images
Bradford School of Optometry and Vision Science, UK
Email:
Classification images (Ahumada, A. J 1996 Perception 26 18) are a technique for deriving the ‘perceptual templates’ used by human observers, on the assumption that observers are linear cross-correlators. In a re-analysis of some recent data (McIlhagga, W 2021 Vision Research 182 36–45), I find that the classification images for detection experiments give a remarkably poor account of how observers detected the stimuli, contra RF Murray et al. (2005 Journal of Vision 5(2)). The implied psychometric functions for the classification images are not steep enough, they are far too efficient, and their internal noise is far too high, compared to humans.
Classification images work better as explanations of human performance when efficiency is reasonably high; in the above study however human efficiency was remarkably low. Here I present some different models, including rectified templates, convolutional templates and squaring, together with open software for fitting and evaluating them. Most of these models provide better accounts of human detection in this experiment than classification images, but important discrepancies remain. I believe that an explanation of such data remains an open question, and good models of trial-by-trial human performance in many cases remain to be discovered.
Sequential judgements of spatial position reveal flexible prior acquisition strategies
University of Nottingham, UK
Email:
Numerous studies have shown that observers incorporate prior knowledge when making judgements about noisy sensory stimuli. However, our understanding of how priors are formed and updated over time remains relatively limited. Here we ask whether prior acquisition strategies adapt to changes in the temporal structure of the stimulus sequence. Observers completed a ‘catch the coin’ task, requiring them to estimate the spatial location of a hidden target based on a noisy visual cue provided on each trial. In half of the blocks of trials, the position of the target position changed gradually across trials, following a random walk sequence. In remaining blocks, we generated a set of target positions using the same random walk procedure, but randomly shuffled their order of presentation. In keeping with previous research, we found that observers’ estimates of target position were biased towards locations on previous trials, an effect that varied systematically with the uncertainty of the cue information provided on each trial. Model comparison revealed clear evidence for flexibility in prior acquisition strategy employed. Observers’ spatial judgements during random walk sequences were most consistent with the weighted combination of the current cue location and the target's position on the previous trial. In contrast, judgements during fully randomised sequences were better characterised by the use of priors formed by averaging target positions across all previous trials. These results indicate that for judgements of spatial position, observers can quickly alter their prior acquisition strategy to match the degree of temporal dependency (autocorrelation) in sensory inputs.
Cognitive-perceptual traits influence the application of physics rules of motion to enhance visual tracking
1King's College London, UK
2Dorset Healthcare University NHS Foundation Trust, UK
3Bournemouth University, UK
Email:
Humans track objects moving smoothly along straight lines excellently. When curves are introduced, performance depends on predictability and individual differences emerge. Schizophrenic patients and high schizotypy individuals show tracking deficits associated with impaired prediction. To shed light on this, we used a paradigm in which 48 healthy participants tracked a moving ball projected across the screen under three gravity direction conditions: gravity, inverted (upwards) and a control (horizontal). Participants completed two inventories quantifying levels of schizotypy (Schizotypal Personality Questionnaire, SPQ) and Autism Spectrum Disorder (Subthreshold Autism Trait Questionnaire, SATQ) traits. A strong correlation of r = 0.735 was measured between the scores. Tracking was best under the gravity condition and worst under the inverted gravity condition. Response dynamics in the gravity condition reveal an early, anticipatory vertical component from stimulus onset which requires participants to apply physics rules to predict motion. More variable delayed anticipatory tracking was measured under the inverted gravity condition consistent with previous findings. We use principal component analysis to unpack the multivariate relationship between trait and eye tracking measures. We identify a distinct component associating the two trait measures with little link to eye tracking, another capturing general eye tracking performance and remainders relating gravity conditions to sub-dimensions within the traits. Findings suggest that tracking deficits resulting from a failure to exploit physics rules are specific to the positive dimension of schizotypy and both rigidity and oddness dimensions of the SATQ. Tracking deficits may be part of a cognitive-perceptual trait encompassing disrupted sensory processing and atypical social interaction.
Seated resting heart rate is associated with temporal visual perception and temporal audio-visual integration in older adults
1Trinity College Dublin, Ireland
2The Irish Longitudinal Study on Ageing, Trinity College Dublin, Ireland
3University College Cork, Ireland
4Mercer Institute for Successful Ageing, St James Hospital, Ireland
Email:
The precision of temporal multisensory integration declines in healthy ageing, a process shaped by both bottom-up (e.g. sensory function) and top-down (e.g. cognition) mechanisms (Hirst et al., 2019 Scientific Reports 9 1–12; Hirst et al., 2022 Aging Brain 2 1–11). There are known links between the autonomic and central branches of the nervous system; for example, cardiovascular function can influence sensory processing and cognition (O’ Callaghan & Kenny, 2016 Yale Journal of Biology and Medicine, 89, 59–71; Skora et al., 2022 Neuroscience & Biobehavioural Reviews 137 1–10). However, it is unknown to what extent cardiovascular function is associated with multisensory integration in ageing. A large sample of older adults (N = 3,232), drawn from The Irish Longitudinal Study on Ageing (TILDA), partook in assessments of unimodal visual and auditory temporal discrimination as well as the Sound Induced Flash Illusion. Seated resting heart rate (RHR) was recorded before these tasks. Faster seated RHR was associated with an improved ability to discriminate two flashes but not two beeps. A faster RHR was also associated with increased SIFI susceptibility when the audio-visual stimuli were presented close together in time (70 ms), but not at longer asynchronies. Our findings suggest a significant association between cardiovascular activity and both visual temporal discrimination and audio-visual temporal integration in older adults, thereby providing evidence for a link between cardiovascular function and sensory perception in ageing.
Surface attitude judgements in synthetic textures and natural images: a method evaluation
1Aston University, UK
2York University, Canada
3Southampton University, UK
Email:
Our knowledge of surface attitude perception derives largely from studies using synthetic stimuli and gauge figure probes proving perspective cues similar to those in the surfaces being judged. Here we assess whether these results generalize to real-world scenes and whether misperception of the gauge figure could produce results that do not reflect human biases. Participants evaluated planar surfaces painted with synthetic textures, and real-world image patches from the SYNS dataset (Adams et al., 2016, Scientific Reports, 6, 35805). Stimuli were viewed monocularly through a 12° aperture. In separate sessions, participants reported surface attitude using a gauge figure and a dial with separate indicators for slant and tilt. Slant: For synthetic textures, participants overestimated slant when using the gauge figure. Although the dial method produced unbiased estimates on average, high slant was overestimated, and low slant underestimated. For real-world surfaces, both methods underestimated slant, but gauge figure settings correlated more strongly with ground truth. Tilt: For synthetic textures, tilt judgements were precise and accurate for both methods. Observers often misinterpreted real-world surfaces as facing left or right, although cardinal facing surfaces were judged correctly. Both the gauge figure and dial method are able to produce reasonable judgement of slant and tilt for synthetic and real-world images. The judgement of tilt in real-world images indicated that observers probably reconstruct allocentric coordinates and misassign surfaces to face left and right, which requires further investigation.
Can we tell the time with microsaccades?
University of Nottingham, UK
Email:
The rate at which we make microsaccades (or fixational saccades) initially decreases and then rebounds following the sudden onset of a visual stimulus. We have previously shown that the amplitude of this rate signature tracks the perceived stimulus contrast and can be used to predict an individual's contrast sensitivity. Here we leverage contextual biases in duration estimation to test whether the dynamics of the rate signature follows physical or perceptual timing. Participants were asked to reproduce intervals defined by the delay between two transient visual stimuli presented sequentially at fixation. Both eyes were tracked using an Eyelink 1000 eye-tracker and microsaccades were detected using an established algorithm. Interval durations either remained constant during a block (800, 1200 or 2000 ms), or were sampled from short (800–1200 ms) or long (1200–2000 ms) log-uniform distributions. When interval timing varied within a block, participants’ reproductions exhibited compressive biases, consistent with the influence of a perceptual prior centred at the mean of the distribution. Comparison of microsaccade rate signatures produced by pairs of stimuli revealed that their timing consistently followed the physical, rather than perceived, temporal separation. However, changes in microsaccade rate before the second stimulus in each pair varied systematically across conditions, consistent with attenuation of eye movements timed to synchronise with the expected interval duration. These findings indicate that patterns of eye movements may accurately signal both the physical timing of external stimuli and internal predictions linked to perceptual distortions.
Software based solution to improve colour rendering accuracy
University of Latvia, Latvia
Email:
Most of computerized colour vision tests available today are displayed on monitors with a resolution of 10 bits per colour channel, although majority of colour monitors used on daily basis have a resolution of 8 bits per colour channel. Despite of most display suboptimal colour resolution for colour vision assessment, in recent years significant advancements of physical resolution are made. We suggest software-based solution to improve colour resolution by exploiting display physical resolution, that is, proposed solution is based on additive colour mixing principle. Method suggested here is intended to improve colour resolution for colour vision assessment methods based on chromatic sensitivity evaluation. To optimize process of finding R,G,B values, which provide best fit between desired x, y, Y values, lookup tables are calculated and sorted for each chromatic direction to be included in colour vision test. Closest fit between x, y, Y values in lookup tables and desired x, y, Y values is found by modified binary search algorithm. Binary search algorithm is designed to provide two sets of R, G, B values that ensures closest fit between desired x, y, Y values. We suggest splitting pixel matrix in groups 2 × 2 pixels and estimate, if best fit is obtained in case 25%, 50% or 75% of pixels within 2 × 2 group are assigned to one of above-mentioned sets of R, G, B values and rest of pixels to remaining set of R, G, B values. The effectivity of the method has been confirmed by theoretical calculations and by spectroradiometer measurements.
A visual correlate of the sustained melanopsin response in humans
University of Manchester, UK
Email:
The human retina contains five photoreceptor classes; rods, cones(3) and intrinsically sensitive melanopsin containing retinal ganglion cells (ipRGCs). The electrophysiological response of ipRGCs is characterised by a sustained activity under prolonged light exposure. We set out to determine whether there might be a visual correlate to this typical ipRGC light response. We decided to study the effect of melanopsin stimulus contrast on a phenomenon known as Troxler fading (a stimulus fades into the background during fixation, indicating retinal/cortical adaptation processes). Using our multi projector system (providing control over four primaries), we presented stimuli that were either high or low in melanopsin contrast and determined the fading latency for both. We found that when a stimulus is rendered at a high melanopsin contrast, it fades much slower into the background than when the melanopsin contrast is lower. These results show that the sustained melanopsin response observed in electrophysiological recordings may have a visual correlate that can be revealed using a Troxler fading paradigm.
Virtual reality as a tool to explore the link between arts and science
1Royal Holloway University, UK
2National Gallery, UK
The collaboration of arts scholars and visual scientist is entering into a new era of cross-disciplinary interaction and use of new technology to explore diverse, difficult and exciting questions about the experience of traditional and contemporary artwork. We started using VR models that can bring ‘reconstructed’ artworks into the lab for detailed behavioural experiments, including eye movements, that can be compared to behaviour in the museum or gallery (e.g. Gulhan, Durant & Zanker, 2021 Scientific Reports 11:18913). In particular, it opens the opportunity to study the active interaction with artworks in 3-dimensiona space, highlighting the skills and intentions of the artists. The interaction between real and illusory elements incorporated in a large, but confined space will be demonstrated for the Carafa Chapel (Santa Maria sopra Minerva in Rome, Filippina Lippi, 1493), which presents an intriguing interaction between physical and painted structures, real and illusory depth, and different levels of narrative. We will show (1) some examples of recordings from participants wearing mobile eye trackers whilst moving freely through the physical chapel in Rome, when exploring areas of the architectural and painted features that they find attractive, intriguing or puzzling; and (2) the initial steps of building a three-dimensional reconstruction of the complete chapel using a 3D laser scanning (LIDAR) of this environment, currently based on photogrammetry. The final VR model will integrate the scanner point-clouds, and will allow us to record fixation patterns to comprehensively study the exploration of Lippi's work in space, 500 years after its creation.