This paper proposes a novel scale-in-scale network (SIS-Net) for ship detection in synthetic aperture radar (SAR) images. SIS-Net establishes multi-scale feature extraction blocks among the single-scale layer-wise level from the backbone network. These blocks can enrich the multi-scale SAR ship feature representation ability. Equipped with the classical feature pyramid network (FPN), SIS-Net can achieve advanced detection performance. Finally, compared with the baseline ResNet-101 backbone network, SIS-Net can improve the AP metric by ~5% on the open SAR ship detection dataset (SSDD).