Decoding object categories from electroencephalography during free viewing reveals early information evolution compared to passive viewing

Object processing is central to visual perception, and its neural substrates inform cognitive and computational models of vision. Electroencephalographic (EEG) research has increasingly adopted multivariate pattern analysis (MVPA) to decode distributed neural representations, moving beyond average signal changes and predefined electrodes. Yet, most MVPA studies use passive viewing paradigms that constrain eye movements and evoke brain activity by external stimuli presentation, creating an artificial dissociation between perception and action. Here, we examine object category representations during active vision by combining MVPA with a deconvolution method for overlapping fixation-related responses. By decoding category information from single fixations, we compare the temporal dynamics of object processing between free viewing (active) and rapid serial visual presentation (passive) conditions. We find that categorical information, across multiple levels of abstraction, can be decoded from individual fixations during natural viewing, and that object representations generalize across passive and active conditions. Notably, representational dynamics time-locked to fixation onset emerge earlier than those time-locked to stimulus onset. We discuss the possibility that object processing begins parafoveally, aligned with saccade onset, rather than foveally at fixation. Thus, information about what and when to expect new information, available in free viewing, facilitates processing. These findings underscore the importance of studying vision in naturalistic settings and highlight key methodological considerations for combining MVPA with deconvolution in overlapping neural data. Object processing is fundamental to visual perception and to cognitive and computational models of vision. Although MVPA has advanced EEG research by enabling the decoding of distributed neural representations with high temporal precision, it has been applied primarily in passive viewing paradigms that constrain natural eye movements. The current study bridges the gap between passive and active vision research by combining MVPA with overlap deconvolution to decode object category information from EEG during natural fixations in free viewing. This approach reveals that category-specific representations emerge earlier in active viewing, likely reflecting processing that begins during saccade planning or execution. The findings underscore the importance of studying vision under ecologically valid conditions and introduce a novel methodological framework for capturing neural dynamics during active perception.