Biology subjects

Wurm, M. F.

Publications and source records attributed to Wurm, M. F..

6 recordsLinked to original sources

bioRxiv2025

Decoding the neural stages from action and object recognition to mentalizing

Higher-level action interpretation, such as inferring underlying intentions and predicting future actions, requires the integration of conceptual action information (e.g. "opening") with semantic knowledge about persons and objects (e.g. "my friend Anna", "pizza box"). However, how the neural systems for action and object recognition and memory interact with each other to form the basis for inferring higher-level mental states remains unclear. Here we use fMRI-based crossmodal multiple regression representational similarity analysis in human female and male participants to elucidate the processing stages from basic action and object recognition to mentalizing. We show that inferring intentions from observed actions or written sentences involves a modality-general network of lateral and medial frontoparietal and temporal brain regions associated with conceptual action and object representation and mentalizing. The representational profiles in these regions are explained by models capturing different types of conceptual information, revealing distinct but partially overlapping networks for action, object, and mental state representation. There was no strict separation of networks for action, object, and mental state representations, arguing against a sequential bottom-up hierarchy from action and object understanding pathways to the mentalizing network. Rather, left-hemispheric regions, specifically ventrolateral prefrontal, inferior parietal and anterior lateral occipitotemporal cortex, showed strong representational overlap, pointing towards a core network for making meaning of action-object structures at a conceptual level. We argue that this core network represents a distributional semantic hub between classic networks for action and object understanding and the mentalizing network. Significance StatementHow does the human brain integrate information from actions, e.g., "open a pizza box", to understand the actions underlying intentions? To do so, the brain needs to combine information from different neural networks--for action and object recognition--and pass them to the mentalizing network for inferring intentions, such as "satisfying hunger". We characterize the interplay of networks using fMRI-based crossmodal multivariate analyses and find that a left-lateralized core network in inferior frontal and parietal cortex and lateral occipitotemporal cortex represents all critical ingredients--conceptual action and object information as well as higher-level mental state representation simultaneously in an overlapping manner. This suggests that this core network is essential for semantic interpretation and functions as bridge between recognition pathways and the mentalizing system.

neuroscience↗

bioRxiv2025

Action prediction relies on high-level holistic and low-level kinematics priors

Biological motion perception is an essential skill that allows us to quickly infer how others move. While this is often cast as inherently predictive process, it remains unclear to what extent different priors shape neural processing of biological motion. We investigated this by comparing the dynamic representational geometry of human magnetoencephalography (MEG) activity and observed biological motion stimuli by means of dynamic representational similarity analysis (dRSA). Under normal viewing conditions, neural representations were indeed predictive and followed an inverse hierarchy: high-level viewpoint-invariant body motion representations were visible before representations of viewpoint-dependent body motion and low-level visual features. Disrupting holistic priors by turning videos upside down selectively reduced high-level predictions. Instead, disrupting kinematics priors by temporal piecewise scrambling eliminated all motion prediction, with neural activity merely reacting to, rather than predicting, visual input. These findings reveal how low- and high-level priors jointly shape predictive neural processing of observed biological motion. Impact statementDisrupting holistic priors selectively interferes with neural prediction of observed biological motion

neuroscience↗

bioRxiv2023

Decoding the physics of observed actions in the human brain

Recognizing goal-directed actions is a computationally challenging task, requiring not only the visual analysis of body movements, but also analysis of how these movements causally impact, and thereby induce a change in, those objects targeted by an action. We tested the hypothesis that the analysis of body movements and the effects they induce relies on distinct neural representations in superior and anterior inferior parietal lobe (SPL and aIPL). In four fMRI sessions, participants observed videos of actions (e.g. breaking stick, squashing plastic bottle) along with corresponding point-light-display stick figures, pantomimes, and abstract animations of agent-object interactions (e.g. dividing or compressing a circle). Cross-decoding between actions and animations revealed that aIPL encodes abstract representations of action effect structures independent of motion and object identity. By contrast, cross-decoding between actions and point-light-displays revealed that SPL is disproportionally tuned to body movements independent of visible interactions with objects. Lateral occipitotemporal cortex (LOTC) was sensitive to both action effects and body movements. Moreover, cross-decoding between pantomimes and animations suggests that right aIPL and LOTC represent action effects even in response to implied object interactions. These results demonstrate that parietal cortex and LOTC are tuned to physical action features, such as how body parts move in space relative to each other and how body parts interact with objects to induce a change (e.g. in position or shape/configuration). The high level of abstraction revealed by cross-decoding suggests a general neural code supporting mechanical reasoning about how entities interact with, and have effects on, each other.

neuroscience↗

bioRxiv2022

A shared neural code for the physics of actions and object events

Observing others actions recruits frontoparietal and posterior temporal brain regions - also called the action observation network. It is typically assumed that these regions support recognizing actions of animate entities (e.g., person jumping over a box). However, objects can also participate in events with rich meaning and structure (e.g., ball bouncing over a box). So far, it has not been clarified which brain regions encode information specific to goal-directed actions or more general information that also defines object events. Here, we show a shared neural code for visually presented actions and object events throughout the action observation network. We argue that this neural representation captures the structure and physics of events regardless of animacy. We find that lateral occipitotemporal cortex encodes information about events that is also invariant to stimulus modality. Our results shed light onto the representational profiles of posterior temporal and frontoparietal cortices, and their roles in encoding event information.

neuroscience↗

bioRxiv2022

Predictive neural representations of naturalistic dynamic input

Adaptive behavior such as social interaction requires our brain to predict unfolding external dynamics. While theories assume such dynamic prediction, empirical evidence is limited to static snapshots and indirect consequences of predictions. We present a dynamic extension to representational similarity analysis that uses temporally variable models to capture neural representations of unfolding events. We applied this approach to source-reconstructed magnetoencephalography (MEG) data of healthy human subjects and demonstrate both lagged and predictive neural representations of observed actions. Predictive representations exhibit a hierarchical pattern, such that high-level abstract stimulus features are predicted earlier in time, while low-level visual features are predicted closer in time to the actual sensory input. By quantifying the temporal forecast window of the brain, this approach allows investigating predictive processing of our dynamic world. It can be applied to other naturalistic stimuli (e.g., film, soundscapes, music, motor planning/execution, social interaction) and any biosignal with high temporal resolution.

neuroscience↗

bioRxiv2019

Exploitation of local and global information in predictive processing

While prediction errors have been established to instigate learning through model adaptation, recent studies have stressed the role of model-compliant events in predictive processing. Specifically, so-called checkpoints have been suggested to be sampled for model evaluation, particularly in uncertain contexts.\n\nUsing electroencephalography (EEG), the present study aimed to investigate the interplay of such global information and local adjustment cues prompting on-line adjustments of expectations. Within a stream of single digits, participants were to detect ordered sequences (i.e., 3-4-5-6-7) that had a regular length of five digits and were occasionally extended to seven digits. Across experimental blocks, these extensions were either rare (low irreducible uncertainty) or frequent (high uncertainty) and could be unexpected or indicated by incidental colour cues.\n\nExploitation of local cue information was reflected in significant decoding of cues vs non-informative analogues using multivariate pattern classification. Modulation of checkpoint processing as a function of global uncertainty was likewise reflected in significant decoding of high vs low uncertainty checkpoints. In line with previous results, both analyses comprised the P3b time frame as an index of excess model-compliant information sampled from probabilistic events.\n\nAccounting for cue information, an N400 component was revealed as the correlate of locally unexpected (vs expected) outcomes, reflecting effortful integration of incongruous information. Finally, we compared the fit of a global model (disregarding local adjustments) and a local model (including local adjustments) using representational similarity analysis (RSA). RSA revealed a better fit for the global model, underscoring the precedence of global reference frames in hierarchical predictive processing.

neuroscience↗