publications

2024

1. Predictive processing in biological motion perception: Evidence from human behavior

   Hüseyin O. Elmas , Sena Er, Ada D. Rezaki , Aysesu Izgi , Buse M. Urgen , Huseyin Boyaci , and Burcu A. Urgen

   Feb 2024

   Abs bioRxiv

   Abstract Biological motion perception plays a crucial role in understanding the actions of other animals, facilitating effective social interactions. While foundation of biological motion perception is rooted in bottom-up processes, as evidenced by point-light display studies, real-world complexities necessitate the involvement of top-down processes, such as attention and expectation. This study investigates the impact of expectations on biological motion perception using a cued individuation task with point-light display stimuli. We conducted three experiments, each providing advance information about distinct aspects of the subsequent biological motion stimuli – specifically information about action, emotion, and gender. Our results revealed a pattern in the action experiment, where participants demonstrated significantly slower response times for incongruent trials than congruent ones, but only under the 75% cue validity condition. This effect was notably absent in the emotion and gender experiments. Our exploration underscores the multi-faceted nature of biological motion perception, highlighting that while the brain adeptly harnesses prior cues to anticipate and interpret stimuli, the nature and reliability of these cues play a pivotal role on their effects. Specifically, action-related information stands out as an important modulator, possibly due to its evolutionary significance and unique neural processing pathway. These findings not only agree with the principles of predictive processing but also pave the way for future research, emphasizing the need to utilize naturalistic, complex stimuli together with neuroimaging methods to create more comprehensive models of biological motion perception.

2022

1. The effect of prior knowledge on biological motion perception

   Hüseyin O. Elmas , Berfin Aydın , Sena Er, Ada D. Rezaki , Ayşesu İzgi , Buse M. Urgen , Huseyin Boyaci , and Burcu A. Urgen

   Journal of Vision, Dec 2022
2. Temporal characteristics of visual processing of agents in action observation

   Hüseyin O. Elmas , Sena Er, Ayse P. Saygin , and Burcu A. Urgen

   In Perception , Dec 2022

   Abs PDF

   To investigate the effect of prior knowledge and presentation mode on the temporal characteristics of visual proces- sing of agents during action observation, we conducted two EEG experiments in which human participants were shown videos and images of 3 agents (robot, human, android) per- forming 8 actions. The main distinction between these experiments was participants’ knowledge about the agent identities before the experiment (called naive or prior experiments). We performed time-resolved representational similarity analysis and modeling to reveal when agent information becomes available in EEG during action observation. To this end, we created a categorical Agent model as well as an Action model and two low-level visual models; and performed multiple regression analysis at each time point. We found that the timing of the agent information depends on the experiment type and the presentation mode. In the naive experiment, agent information was avail- able for a longer duration during the processing of images than videos (66–418ms vs. 90–130ms). In the prior experiment, agent information was present for similar durations in image and video conditions (82–296ms vs. 90–242ms). In summary, our results suggest that prior information and presentation mode modulate the onset and duration of agent information in EEG data.

2021

1. The effect of prior knowledge on visual processing of agents: Time-resolved representational similarity analysis on EEG data

   Sena Er, Huseyin O. Elmas , Ayse P. Saygin , and Burcu A. Urgen

   Dec 2021

   Abs

   Recognition of agents during action observation is a crucial skill for many species including humans. Visual processing of agents is considered to take place in the pSTS node of the Action Observation Network; however, its temporal course and the factors that affect it still remain unknown. We investigated how prior knowledge about agents affects temporal characteristics of neural processing. We conducted 2 EEG experiments with human subjects. In both experiments (12 female, 15 male in total), subjects were presented videos and still images of 3 agents who performed 8 different actions as EEG was recorded. In the first experiment (Prior experiment), subjects (n=19) were familiarized with the stimuli and informed about the identity of the agents prior to the experiment. In the second experiment (Naive experiment), subjects (n=16) were naive to the identity of the agents. After standard preprocessing, we linked the EEG data in a time-resolved manner to a categorical Agent Model using representational similarity analysis (Kendall Tau correlation). We also used 3 additional models that capture the non-target variables within presented stimuli, a categorical Action model and two low-level visual models: pixel-wise intensity and pixel-wise motion energy. Finally, we did multiple regression analysis at each time point to determine how much each individual model could independently predict the neural data; thus, enabling us to regress out the effect of non-target variables on the correlation between the Agent Representational Dissimilarity Matrix (RDM) and the neural RDM (Bonferroni corrected for all time points and conditions, p<0.05). Our results show that the most informative channels about agent information were the ones over the occipital and parietal cortex. In those channels, there was a specific time range when agent information was available but it depended on whether the subjects were naive about the identity of the agents and the presentation mode. The comparison between Prior and Naive experiments showed that when subjects had no prior knowledge about agents, agent information became available for a longer duration during processing of still images (66-418 ms) than videos (90-130 ms). However, when subjects had prior knowledge about the agents, agent information was present for similar durations in both stimulus presentation modes: 82-196 ms for still images or 90-172 ms for videos. We conclude that prior information affects temporal characteristics of visual processing of agent perception by modulating the onset and duration of when agent information is available.

2020

1. The effect of prior knowledge on visual processing of agents: time-resolved representational similarity analysis on EEG data

   Huseyin O. Elmas , Sena Er, Ayse P. Saygin , and Burcu A. Urgen

   In Anatomy: International Journal of Experimental & Clinical Anatomy , Nov 2020

   Abs

   Objective: Visual processing of actions involves identification of the agent who performs the action and the type of action. In the presented study, we investigated how prior knowledge affects temporal characteristics of neural processing during perception of videos and still images that include various agents and actions. Methods: We conducted two EEG experiments with human participants. In both experiments, participants were presented videos and images of 3 agents performing 8 actions while EEG was recorded. In the first experiment (Prior), participants (n=19) were familiarized with the stimuli and informed about the identity of the agents before the experiment. In the second experiment (Naive), participants (n=16) were naive to the identity of the agents. EEG data was preprocessed with standard procedures including filtering, referencing, and artifact rejection. To investigate the time course of agent and action processing, we linked the EEG data in a time-resolved manner to two categorical models, Agent and Action models, using representational similarity analysis. We then performed multiple regression analysis by including Agent and Action models as well as two low-level stimulus feature models as predictors to reveal the unique contribution of Agent and Action models to EEG data. Results: Our regression analysis with RSA shows that when subjects had no prior knowledge about agents, agent information was available for a longer duration during processing of images (66-406ms) than videos (92-128ms) (pcO.Ol, Bonferroni corrected for all time points). However, when subjects had prior knowledge about the agents, the difference between the still images and videos was diminished. We did not find any time interval in EEG data when Action information was available (pcO.Ol, Bonferroni corrected for all time points). Conclusion: In conclusion, prior information affects temporal characteristics of visual processing of agent perception by modulating the onset and duration of when agent information is available.

2. Eylemlerin Ve Aktörlerin Görsel İşlenmesinde Ön Bilginin Etkisi: EEG Verisinde Zaman Boyutunda Temsil Benzerliği Analizi

   Hüseyin O. Elmas , Sena Er, Ayse P. Saygin , and Burcu A. Urgen

   Nov 2020
3. Temporal characteristics of visual processing of actions: Time-resolved RSA on EEG data

   Hüseyin O. Elmas , Sena Er, Ayse P. Saygin , and Burcu A. Urgen

   Nov 2020