Neural Representations Across Multiple Stages in a Go/No-Go Task: Evidence From Multivariate Pattern Analysis

The Go/No-Go task is one of the most commonly used paradigms in psychological research, yet traditional studies rely on single event-related potential (ERP) components, limiting insights into multi-stage temporal dynamics. This study recorded electroencephalography (EEG) data from 72 participants performing a Go/No-Go task, and analyzed the data using univariate ERP analysis and multivariate pattern analysis (MVPA), including decoding, temporal generalization, and weight projection analyzes. Conventional ERP analysis replicated larger N2 and P3 components for No-Go trials relative to Go trials. Critically, decoding analysis revealed that Go and No-Go trials were discriminable as early as ~100 ms post-stimulus. Temporal generalization analysis identified distinct phases: early temporally generalizable representations (~300-400 ms), a highly time-specific stage (~400-700 ms), and later temporally generalizable representations (~700-1700 ms), suggesting a multi-stage evolution of neural representations during the task. Weight projection analysis highlighted the involvement of occipital, prefrontal, and parietal regions, indicating the recruitment of diverse neural networks throughout the task. Additionally, post-trial history effects demonstrated a temporal crossover: early processing (before ~100 ms) benefited from prior-trial information, whereas later stages (320-1500 ms) were dominated by current-trial information. These findings provide new insights into how the Go/No-Go task is supported by a multi-stage, dynamically reorganized neural code that is continuously shaped by recent trial history.