In recent years, the application of error-related potentials (ErrPs) in brain-computer interfaces (BCIs) has expanded the amount of information available for decoding user intent. To extract users’ preferences of BCI control outcomes from ErrP across a wide range of control scenarios, researchers seek to rapidly deploy ErrP decoding models to new BCI paradigms. However, variability in ErrP waveform characteristics across paradigms limits the direct application of pre-trained classifiers in novel paradigms. Existing transfer learning approaches for cross-paradigm ErrP decoding lack a necessary physiological foundation and therefore cannot provide physiologically meaningful reference-state data for transfer learning. To address this limitation, we propose that EEG responses recorded under a neutral condition which is free from task-related expectations can be utilized as physiology-informed reference data for cross-paradigm transfer learning. Significant cross-paradigm correlations between this neutral condition responses and ErrP responses in both the latency and amplitude of waveforms support the reference value of this neutral condition. Furthermore, we introduce a Neutral Condition Aligned Training Protocol (NCTP) for cross-paradigm ErrP data alignment. The significant improvements in the balanced accuracy of cross-paradigm decoding demonstrate the effectiveness of this NCTP approach.