Biology subjects

Herzog, M. H.

Publications and source records attributed to Herzog, M. H..

12 recordsLinked to original sources

bioRxiv2025

Higher inter-trial latency variability contributes to reduced visual EEG responses in schizophrenia

Patients with schizophrenia show strong impairments in visual backward masking, which are associated with reduced EEG N1 responses. However, it is currently unclear whether reduced N1 amplitudes in patients reflect attenuated neural responses or increased inter-trial latency variability, as both can lead to reduced trial-averaged responses. Previous studies using trial-averaged data cannot distinguish between these two possibilities. Here, we estimated inter-trial latency variability of the visual N1 component and found significantly increased variability in patients compared to controls. Inter-trial latency-variability was a strong predictor of the N1 amplitude in both groups. Importantly, after accounting for the effect of latency variability in the group comparison, patients continued to exhibit significantly reduced N1 amplitudes, although the effect size diminished from large to medium. These findings indicate that both higher latency variability and attenuated neural responses contribute to visual processing deficits in schizophrenia.

neuroscience↗

bioRxiv2025

Mid-level visual features engage V4 to support sparse object recognition under uncertainty

Cortical visual prostheses aim to restore sight by stimulating the visual cortex, yet current approaches remain limited. Implants targeting V1 typically evoke isolated phosphenes that do not support coherent object recognition. We asked whether mid-level visual features, specifically curve segments associated with area V4, could provide a more efficient substrate for artificial vision. Using fMRI, 24 participants performed an object recognition task with stimuli fragmented into either phosphene-like dots or curve segments at varying densities. Behaviorally, curve segments enabled recognition at lower density thresholds and produced steeper psychometric slopes. Neurally, both fragment types engaged V1, but only curve segments recruited V4, and correct recognition of segments relative to phosphenes evoked broader occipito-temporo-parietal responses. Multivariate decoding showed that successful recognition of phosphenes was predicted by early visual and cingulate activity, whereas recognition of segments was predicted only by V4, indicating a shift toward mid-level representational support. Mediation analyses further showed that phosphene-based recognition relied on anterior cingulate cortex under conditions of higher uncertainty, while segment-based recognition depended primarily on visual cortex. Together, these results show that mid-level fragments confer perceptual and neural advantages, reducing reliance on both dense stimulation and downstream uncertainty resolution, and identify V4 as a promising target for next-generation cortical prostheses.

neuroscience↗

bioRxiv2024

Neural correlates of unconscious and conscious visual processing

In some forms of postdictive phenomena, later events influence the perception of earlier ones, suggesting that conscious perception may be preceded by extended periods of unconscious processing. An example is the Sequential Metacontrast (SQM) paradigm, in which vernier offsets are unconsciously integrated over several hundred milliseconds before conscious perception. Obviously, the integrated percept can only emerge after each individual element in the stream has been processed. Thus, the SQM provides a unique opportunity to dissociate unconscious from conscious stages of visual processing, as these stages are well separated in time. Using EEG during the SQM, we identified two distinct stages of neural activity: an early occipital EEG activity pattern ([~]200 ms after the initial vernier) associated with unconscious processing, and a later centro-parietal EEG pattern ([~]400-600 ms after SQM onset) associated with the integrated percept and the behavioral report. We propose that the transition between these neural patterns marks the shift from unconscious encoding of individual visual stimuli to their integrated percept.

neuroscience↗

bioRxiv2024

Minimizing the number of phosphenes required for object recognition under prosthetic vision

Cortical prostheses offer the potential for partial vision restoration in individuals with blindness by stimulating V1 neurons to produce phosphenes. However, the low number of phosphenes that can be elicited in practice makes encoding of whole objects difficult, and the round shape of phosphenes lack the contour cues necessary for perceptual grouping. We propose a minimalistic encoding approach that focuses on essential visual information. We fragmented objects contours into either phosphenes or curved segments, providing either low or high local visual information. 46 participants identified these fragmented objects in a free-naming task. The number of fragments gradually increased to quantify the minimum number of phosphenes and segments necessary to recognize objects. Most objects could be recognized with only 65 phosphenes, which is in the range of implantable electrodes in human patients. Participants required 27% fewer segments than phosphenes to recognize objects. Including individual objects as a random effect in a linear mixed model substantially increased the explained variance, suggesting that the minimal number of fragments required for object recognition in prosthetic vision strongly depends on the particular object. Our results demonstrate that a minimalistic approach can substantially reduce the number of phosphenes required for recognition, emphasizing the importance of identifying critical object features to minimize brain stimulation in visual prostheses.

neuroscience↗

bioRxiv2023

Is there a neural common factor for visual illusions?

It is tempting to map interindividual variability in human perception to variability in brain structure or neural activity. Indeed, it has been shown that susceptibility to size illusions correlates with the size of primary visual cortex V1. Yet contrary to common belief, illusions correlate only weakly at the perceptual level, raising the question of how they can correlate with a localized neural measure. In addition, mounting evidence suggests that there is substantial interindividual variability not only in neural function and anatomy but also in the mapping between the two, which further challenges the findings of a neural common factor for illusions. To better understand these questions, here, we re-evaluated previous studies by correlating illusion strengths in a battery of 13 illusions with the size of visual areas and population receptive field sizes. We did not find significant correlations either at the perceptual level or between illusion susceptibility and visual functional neuroanatomy.

neuroscience↗

bioRxiv2023

Oscillatory traveling waves reveal predictive coding abnormalities in schizophrenia

The computational mechanisms underlying psychiatric disorders are hotly debated. One hypothesis, grounded in the Bayesian predictive coding framework, proposes that schizophrenia patients have abnormalities in encoding prior beliefs about the environment, resulting in abnormal sensory inference, which can explain core aspects of the psychopathology, such as some of its symptoms. Here, we tested this hypothesis by identifying oscillatory traveling waves as neural signatures of predictive coding. By analyzing a EEG dataset. comprising 146 schizophrenia patients and 96 age-matched healthy controls, we found that schizophrenia patients have stronger top-down alpha-band traveling waves compared to healthy controls during resting state, reflecting stronger precise priors at higher levels of the predictive processing hierarchy. Conversely, we found stronger bottom-up alpha-band waves in schizophrenia patients during a visual task reflecting an alteration of lower sensory priors. Our results yield a novel spatial-based characterization of oscillatory dynamics in schizophrenia, considering brain rhythms as traveling waves and providing a unique framework to study the different components involved in a predictive coding scheme. Altogether, our findings significantly advance our understanding of the mechanisms involved in fundamental pathophysiological aspects of schizophrenia, promoting a more comprehensive and hypothesis-driven approach to psychiatric disorders. SignificanceWe provide novel evidence favoring the Bayesian predictive coding interpretation of schizophrenia. Relying on computational and experimental works that characterized electrophysiological correlates of predictive processes, we investigate oscillatory traveling waves in EEG data of 146 schizophrenia patients and 96 age-matched healthy controls. Our results reveal stronger top-down alpha-band traveling waves in schizophrenia patients, reflecting an increase in the precision of high-level priors. On the other hand, we observed an increase in bottom-up alpha-band waves during a visual task, in line with the proposed reduction in precision of low-level sensory priors. Our findings suggest that traveling waves analysis is a versatile technique to probe predictive processing mechanisms in different cognitive processes. Impairments in this mechanism may underlie perceptual alterations as well as the pronounced clinical symptoms of schizophrenia. The strong hypothesis-driven nature of our results accentuates the relevance of our findings.

neuroscience↗

bioRxiv2023

Intact reinforcement learning in healthy ageing

What does age in ageing? Results in reinforcement learning (RL) are mixed. Some studies found deteriorated performance in older participants compared to younger controls whereas other studies did not. Daniel et al. (2020) suggested that task demand can explain these differences, with less demanding tasks showing no effect of age. Here, we increased the task demand of previous studies turning them into a classic navigation task. We extracted 4 behavioral parameters and 2 parameters (learning and exploration rates) of a classic Q-learning model. Except for one specific parameter, all other parameters showed no group differences, i.e., RL turned out to be intact in older individuals also with higher task demands. It is important to publish such null results to avoid the stigmatizing impression of an overall performance deficit among older people.

neuroscience↗

bioRxiv2022

The curse of optimism: a persistent distraction by novelty

Humans successfully explore their environment to find extrinsic rewards, even when exploration requires several intermediate reward-free decisions. It has been hypothesized that intrinsic rewards such as novelty guide this reward-free exploration. However, different intrinsic rewards lead to different exploration strategies, some prone to suboptimal attraction to irrelevant stochastic stimuli, sometimes called the noisy TV problem. Here, we ask whether humans show a similar attraction to reward-free stochasticity and, if so, which type of intrinsic reward guides their exploration. We design a multi-step decision-making paradigm where human participants search for rewarding states in an environment with a highly stochastic but reward-free sub-region. We show that (i) participants persistently explore the stochastic sub-region and (ii) their decisions are best explained by algorithms driven by novelty but not by optimal algorithms driven by information gain. Our results suggest that humans use suboptimal but computationally cheap strategies for exploration in complex environments.

neuroscience↗

bioRxiv2020

Novelty is not Surprise: Exploration and learning in human sequential decision-making

Classic reinforcement learning (RL) theories cannot explain human behavior in response to changes in the environment or in the absence of external reward. Here, we design a deep sequential decision-making paradigm with sparse reward and abrupt environmental changes. To explain the behavior of human participants in these environments, we show that RL theories need to include surprise and novelty, each with a distinct role. While novelty drives exploration before the first encounter of a reward, surprise increases the rate of learning of a world-model as well as of model-free action-values. Even though the world-model is available for model-based RL, we find that human decisions are dominated by model-free action choices. The world-model is only marginally used for planning but is important to detect surprising events. Our theory predicts human action choices with high probability and allows us to dissociate surprise, novelty, and reward in EEG signals.

neuroscience↗

bioRxiv2020

Electrophysiological correlates of visual backward masking in patients with bipolar disorder

BackgroundIn visual backward masking (VBM), a target is followed by a mask that decreases target discriminability. Schizophrenia patients (SZ) show strong and reproducible masking impairments, which are associated with reduced EEG amplitudes. Patients with bipolar disorder (BP) show masking deficits, too. Here, we investigated the neural EEG correlates of VBM in BP. Methods122 SZ, 94 unaffected controls, and 38 BP joined a standard VBM experiment. 123 SZ, 94 unaffected controls and 16 BP joined a corresponding EEG experiment, analyzed in terms of the global field power. ResultsAs in previous studies, SZ and BP show strong masking deficits. Importantly and similarly to SZ, BP show decreased global field power amplitudes at approximately 200 ms after the target onset, compared to controls. ConclusionsThese results suggest that VBM deficits are not specific for schizophrenia but for a broader range of functional psychoses. Potentially, both SZ and BP show deficient target enhancement.

neuroscience↗

bioRxiv2019

Capsule Networks but not Classic CNNs Explain Global Visual Processing

Classically, visual processing is described as a cascade of local feedforward computations. Feedforward Convolutional Neural Networks (ffCNNs) have shown how powerful such models can be. However, using visual crowding as a well-controlled challenge, we previously showed that no classic model of vision, including ffCNNs, can explain human global shape processing (1). Here, we show that Capsule Neural Networks (CapsNets; 2), combining ffCNNs with recurrent grouping and segmentation, solve this challenge. We also show that ffCNNs and standard recurrent CNNs do not, suggesting that the grouping and segmentation capabilities of CapsNets are crucial. Furthermore, we provide psychophysical evidence that grouping and segmentation are implemented recurrently in humans, and show that CapsNets reproduce these results well. We discuss why recurrence seems needed to implement grouping and segmentation efficiently. Together, we provide mutually reinforcing psychophysical and computational evidence that a recurrent grouping and segmentation process is essential to understand the visual system and create better models that harness global shape computations. Author SummaryFeedforward Convolutional Neural Networks (ffCNNs) have revolutionized computer vision and are deeply transforming neuroscience. However, ffCNNs only roughly mimic human vision. There is a rapidly expanding body of literature investigating differences between humans and ffCNNs. Several findings suggest that, unlike humans, ffCNNs rely mostly on local visual features. Furthermore, ffCNNs lack recurrent connections, which abound in the brain. Here, we use visual crowding, a well-known psychophysical phenomenon, to investigate recurrent computations in global shape processing. Previously, we showed that no model based on the classic feedforward framework of vision can explain global effects in crowding. Here, we show that Capsule Neural Networks (CapsNets), combining ffCNNs with recurrent grouping and segmentation, solve this challenge. ffCNNs and recurrent CNNs with lateral and top-down recurrent connections do not, suggesting that grouping and segmentation are crucial for human-like global computations. Based on these results, we hypothesize that one computational function of recurrence is to efficiently implement grouping and segmentation. We provide psychophysical evidence that, indeed, grouping and segmentation is based on time consuming recurrent processes in the human brain. CapsNets reproduce these results too. Together, we provide mutually reinforcing computational and psychophysical evidence that a recurrent grouping and segmentation process is essential to understand the visual system and create better models that harness global shape computations.

neuroscience↗

bioRxiv2019

Crowding Reveals Fundamental Differences in Local vs. Global Processing in Humans and Machines

Feedforward Convolutional Neural Networks (ffCNNs) have become state-of-the-art models both in computer vision and neuroscience. However, human-like performance of ffCNNs does not necessarily imply human-like computations. Previous studies have suggested that current ffCNNs do not make use of global shape information. However, it is currently unclear whether this reflects fundamental differences between ffCNN and human processing or is merely an artefact of how ffCNNs are trained. Here, we use visual crowding as a well-controlled, specific probe to test global shape computations. Our results provide evidence that ffCNNs cannot produce human-like global shape computations for principled architectural reasons. We lay out approaches that may address shortcomings of ffCNNs to provide better models of the human visual system.

neuroscience↗