Recent discovery of ROS1 gene fusion in a subset of lung cancers has raised clinical interest, because ROS1 fusion–positive cancers are reportedly sensitive to kinase inhibitors. To better understand these tumors, we examined 799 surgically resected non–small cell lung cancers by reverse transcriptase polymerase chain reaction and identified 15 tumors harboring ROS1 fusion transcripts (2.5% of adenocarcinomas). The most frequent fusion partner was CD74 followed by EZR. The affected patients were often younger nonsmoking female individuals, and they had overall survival rates similar to those of the ROS1 fusion–negative cancer patients. All the ROS1 fusion–positive tumors were adenocarcinomas except 1, which was an adenosquamous carcinoma. Histologic examination identified an at least focal presence of either solid growth with signet-ring cells or cribriform architecture with abundant extracellular mucus in 53% of the cases. These 2 patterns are reportedly also characteristic of anaplastic lymphoma kinase (ALK)-rearranged lung cancers, and our data suggest a phenotypic resemblance between the ROS1-rearranged and ALK-rearranged tumors. All tumors except 1 were immunoreactive to thyroid transcription factor-1. Fluorescence in situ hybridization using ROS1 break-apart probes revealed positive rearrangement signals in 23% to 93% of the tumor cells in ROS1 fusion–positive cancers, which were readily distinguished using a 15% cutoff value from 50 ROS1 fusion–negative tumors tested, which showed 0% to 6% rearrangement signals. However, this perfect test performance was achieved only when isolated 3′ signals were included along with classic split signals in the definition of rearrangement positivity. Fluorescence in situ hybridization signal patterns were unrelated to 5′ fusion partner genes. All ROS1 fusion–positive tumors lacked alteration of EGFR, KRAS, HER2, ALK, and RET genes.