 The proposed SCAMLP net is a novel self-supervised learning SSL-based method for EEG-based motor imagery, MI, decoding. It uses a pretext task to capture long-range temporal information in EEG trials, and then applies a MLP mixer to the classification task for signals instead of for images. Additionally, an attention mechanism is integrated into the MLP mixer to adaptively estimate the importance of each EEG channel without any prior information. This allows the model to better learn more long-range temporal information and global spatial features of EEG signals. Experimental results show that the proposed SCAMLP net outperforms other state-of-the-art methods on both the MI2 dataset and the BCI competition for dataset 2A. This article was authored by Yen Ben-Hee, Jiang Lu, Jun Wang, and others.