This paper proposes a novel path aggregation attention feature pyramid network (PAA-FPN) for synthetic aperture radar (SAR) ship detection. PAA-FPN adds a bottom-up branch in the raw FPN to enhance spatial location information. Moreover, in order to enhance information transmission benefits, PAA-FPN is also embedded with an attention module. Experimental results on the first open SAR ship detection dataset (SSDD), PAA-FPN can improve the detection accuracy by ~3%, compared with the baseline FPN method. We also conduct ablation studies to confirm the effectiveness of the attention module.