Related to lines:
 Eye Movements & Perception and Action |
Summary: The consequence of the non-homogeneous distribution of photoreceptors across the human retina is that we make about three saccades per second in order to foveate different parts of the scene for highly detailed information acquisition. Saccades are very fast, and of very short duration, and are always followed by periods of fixation. Surprisingly, despite being extremely brief, extensive research has delved into the relation between saccades and visual perception, whereas fixations, which are often more than twice as long, have received little attention. Fixation periods play a crucial role in perceiving the world around us. During these periods, gaze remains relatively stable, and the visual system engages in processing foveal information, sending it to visual memory areas for scene reconstruction, and obtains peripheral information to, for example, plan the next saccade. Thus, fixation involves processing both foveal and peripheral information, a fundamental aspect for survival. For instance, picture yourself driving a car. When doing so, gaze is normally directed to the road ahead, but simultaneous processing of peripheral information to for example, detect a traffic light shift is essential. Visual sensitivity across the visual field fluctuates across time and varies especially with movement preparation and saccade execution. However, it remains unknown whether and if so, how, human's sensitivity across the visual field is modulated over the course of fixation. This project addresses this fundamental question aiming to achieve two main objectives. The first one is to describe the temporal dynamics of foveal and peripheral processing during fixation; the second one is to determine whether the visual system can prioritize the processing of foveal or peripheral information depending on the needs. To study this, we will use an innovative experimental paradigm based on psychophysical designs that will allow determining the temporal dynamics of foveal and peripheral processing upon saccade landing, during fixation. The results of this project will not only elucidate differences in foveal and peripheral sensitivity across time, during fixation, but will also contribute to create a highly detailed map of the temporal sensitivity to changes across the retina, analogous to visual acuity maps. Furthermore, the project will explore the effects that having to actively control fixation may have on foveal and peripheral processing, and the potential malleability in prioritizing information from different parts of the scene during fixation based on specific requirements. The achievement of our goals will significantly advance our comprehension of how we process visual information, providing specialized insights into the temporal dynamics of human visual perception. Beyond scientific knowledge, the results hold potential implications for diverse fields, including the development of engaging and more natural virtual and augmented reality environments, and the enhancement of machine vision algorithms to simulate human visual perception.
Related to lines:
 Decision Making in Complex Environments & Psychophysics | Eye Movements & Perception and Action |
Optic flow, the dynamic pattern of light projected from the environment to the observer, is pivotal for understanding human interaction with the three-dimensional visual world. When optic flow is projected onto the retinas, it encapsulates motion information about objects and self- motion. This project focuses on the central controversy in optic flow research: the necessity and complexity of recovering metric informationspecific measurements such as position, distance traveled, and velocityfrom the complex gradients on the retinal planes. Traditionally, theories have suggested that metrically accurate three-dimensional environmental information can be derived from optic flow. However, contemporary research highlights a reliance on relative retinal cues and angular variables over precise metric reconstructions. Influenced significantly by eye and head movements, this shift raises questions about our capability to extract metric information from the retinal flow accurately, especially regarding the physical velocity of objects or self-motion.
This project aims to address these challenges using Continuous Psychophysics (CPsych), an innovative approach for measuring perceptual uncertainties and decision-making processes. CPsych, with its continuous response format, offers a more nuanced analysis of sensory processing, making it an ideal tool to explore the implicit uncertainties of visual features. The project unfolds across two objectives.
The first objective will extensively utilize CPsych to determine whether difficulties in extracting metric velocity from local Optic flow stem from inherent complexities in interpreting Optic flow or from a reliance on the perceptual judgments of classical psychophysics. A second objective moves beyond local object motion to global optic flow, focusing on the accurate detection of the focus of expansion (FOE) amidst complex retinal gradients caused by eye movements. Here, the project seeks to improve the precision of understanding optic flow by employing CPsych and machine learning models, simulating FOE motion dynamics that are more akin to natural conditions.
By addressing these objectives, the project aims to make a significant contribution to our understanding of how we perceive and interact with our dynamic 3D environment. The findings have the potential to reshape existing models of visual perception and inform practical applications in various fields, from autonomous navigation to virtual reality, where accurate interpretation of optic flow is crucial.
Related to lines:
| Decision Making in Complex Environments & Psychophysics |
Schizophrenia, a psychiatric disorder affecting approximately 1% of the population, entails substantial personal, familial, and societal costs due to its prevalence and severity. To advance both pharmacological and psychotherapeutic approaches, a comprehensive understanding of the neural and computational alterations underlying the disorder is crucial.
Research indicates that individuals with schizophrenia manifest perceptual anomalies beyond hallucinatory disturbances, suggesting differential processing of visual stimuli compared to healthy individuals. These perceptual alterations may unveil dysfunctions in visual neural circuits, hinting at broader anomalies in the disease's computational mechanisms.
A prominent example of visual processing deficit in schizophrenia is impaired contrast detection, a critical aspect involving the recognition of low-contrast stimuli. Numerous studies have consistently demonstrated reduced contrast detection in individuals with schizophrenia compared to healthy controls.
We have hypothesized that if deficits in contrast perception in schizophrenia arise from glutamatergic hypofunction, they should be particularly pronounced not in detection but in discrimination around low-contrast pedestals and should also be observed in patients with anti-NMDAR encephalitis, a condition characterized by glutamatergic system hypofunction. Preliminary results supported this hypothesis after controlling for task engagement.
To validate and deepen our understanding, we propose to investigate contrast discrimination in a new sample of participants with schizophrenia. Additionally, to discern the origin of the deficit, we will employ hidden Markov models with Bernoulli generalized linear model observations (GLM-HMM) to identify different internal states contributing to observed task performance differences. This approach aims to ascertain whether the deficit is attributed to lower contrast sensitivity in patients, even during highly engaged states, or if it results from patients being less frequently in highly engaged states.
Identifying a deficit in contrast discrimination as a potential biomarker for glutamatergic hypofunction in psychosis holds implications for the field. This discovery could facilitate the categorization of subgroups within schizophrenia, enabling a more targeted approach for those with a more affected glutamatergic system.
Related to lines:
| Decision Making in Complex Environments & Psychophysics |
The present proposal aims at the further development of two models that were developed in my previous research project.
The first model is a modern Hopfield network with dynamic memory. The novelty of this model is that the memory with the stored patterns is dynamically updated. A collision detection model based on the latter network achieved results which are comparable with (and in some cases even superior to) the state-of-the-art approaches for visual collision detection. However, the dynamic Hopfield model in its current state cannot selectively adapt the stored patterns. Therefore I propose to merge the dynamic Hopfield model with Adaptive Resonance Theory (ART). ART is essentially a powerful learning mechanism, which also specifies how bottom-up mechanisms in the brain interact with top-down memory predictions. In this way, I expect to obtain "Hopfield-ART", that is an auto-associative memory which is capable of dynamically adapting the stored patterns in an unsupervised fashion. The Hopfield-ART model therefore would explain how the brain can recognize and retrieve incomplete and/or noisy patterns on the one hand (= auto-associative property), and update the stored patterns with new information on the other (e.g. faces that change with time due to ageing).
Nevertheless, the Hopfield-ART model would still be an abstract model which ignores several features of biological neurons (e.g. spiking, dendrites, synapses, receptors). These features are taken into account by my second model about dendritic learning (DL). The current version of the DL-model shows a rather impressive recognition performance of learned patterns versus control patterns (e.g., real-world images). In the context of predictive coding, the DL-model also clarifies how a neuron can reduce its responses to specific frequencies of spiking, while its responses to previously unseen spike patterns remain largely unaffected.
However, it cannot re-create yet the pattern that it has learned before, such as the brain does with dreaming and visual imagination. This shortcoming of the DL-model will be addressed with the present proposal. Therefore, while Hopfield-ART is more on the articficial intelligence and algorithmic side, the DL-model will make concrete predictions on how a single neuron can store and re-create many patterns. Specifically, both models can be tuned to implement predictive coding, which is, for example, important to suppress background information in collision detection. I plan to use the latter "application" in order to demonstrate the utility of the new models, and possiby
also to understand how the Lobula Giant Movement Detector neuron of the locust processes incoming visual information with its dendrite.
Related to lines:
 Number Processing & Probabilistic Reasoning |
Math anxiety is a state of tension and apprehension that affects some individuals when they have to deal with mathematical stimuli. According to the recent PISA 2022 study, 37% of the examined Spanish teenagers (compared to the European average of 17%) experience math anxiety. This condition leads to avoidance of mathematical stimuli, resulting in poorer education and worse work and economic prospects. Hence, there is a need to study this phenomenon and its relationship with performance in mathematical tasks.
Recent studies, including some from our own group, have suggested that the difficulties of individuals with math anxiety may stem from a deficit in their executive functions, hindering sustained and goal-directed attention. Continuing in this line, this project aims to investigate metacognition in individuals with high math anxiety (HMA). Good metacognition, or thinking about one's own cognition, is related to the quality of executive functions and is crucial for achieving personal goals. Specifically, we will explore two aspects of metacognition:
metacognitive monitoring refers to the subjective evaluation of the success of the ongoing or just completed activity, while metacognitive control refers to the regulation of cognitive activity by initiating or modifying the allocation of available resources based on the results of metacognitive judgments. Throughout six behavioral experiments, we will compare the performance of two extreme groups differing in math anxiety. The initial experiments will determine whether individuals with HMA differ from their peers in the calibration of their
metacognitive judgments, and whether these differences occur both when judgments are measured explicitly and indirectly. Next, we will investigate whether both groups differ in their ability to select the most appropriate strategy based on the characteristics of the problem they face. Subsequently, we will focus on aspects more related to control and investigate whether individuals with HMA show less flexibility in selecting strategies and are less capable of adjusting their decisions based on the results of their previous monitoring. Finally, we will also examine whether individuals with HMA differ from their peers in their ability to manage their cognitive and temporal resources when learning mathematical information. So far, studies on metacognition and math anxiety are scarce and have never approached it from all the aspects that will be addressed in this project. Acquiring relevant information about each of these components of metacognition in individuals with HMA should be the first step in determining which components are preserved and can compensate for the effects of emotion on cognition, and which ones require the design of intervention programs for improvement.