This paper explores the effectiveness of sparse representation and collaborative representation in face recognition (FR). While sparse representation (SRC) has been widely used, it is often emphasized that the $l_1$-norm sparsity plays a crucial role in improving FR accuracy. However, the paper argues that it is actually the collaborative representation (CR) that is more important. CR involves using all training samples from different classes to represent the query sample, which enhances the discrimination and robustness of the classification. The authors propose a new method called Collaborative Representation based Classification with Regularized Least Square (CRC_RLS), which is simpler and more efficient than SRC. Extensive experiments on various face databases, including Extended Yale B, AR, and Multi-PIE, demonstrate that CRC_RLS achieves competitive classification results while being significantly faster. The paper concludes that CR is the key mechanism improving FR accuracy, and CRC_RLS is a promising alternative to SRC.This paper explores the effectiveness of sparse representation and collaborative representation in face recognition (FR). While sparse representation (SRC) has been widely used, it is often emphasized that the $l_1$-norm sparsity plays a crucial role in improving FR accuracy. However, the paper argues that it is actually the collaborative representation (CR) that is more important. CR involves using all training samples from different classes to represent the query sample, which enhances the discrimination and robustness of the classification. The authors propose a new method called Collaborative Representation based Classification with Regularized Least Square (CRC_RLS), which is simpler and more efficient than SRC. Extensive experiments on various face databases, including Extended Yale B, AR, and Multi-PIE, demonstrate that CRC_RLS achieves competitive classification results while being significantly faster. The paper concludes that CR is the key mechanism improving FR accuracy, and CRC_RLS is a promising alternative to SRC.