Lung cancer is the leading cause of cancer-related mortality worldwide. Methods for early detection of lung cancer, such as computerized tomography scanning technology, often discover a large number of small lung nodules, posing a new problem to radiologists and chest physicians. The vast majority of these nodules will be benign, but there is currently no easy way to determine which nodules represent very early lung cancer. Adjuvant testing with PET imaging and nonsurgical biopsies has a low yield for these small indeterminate nodules, carries potential morbidity and is costly. Indeed, purely morphological criteria seem to be insufficient for distinguishing lung cancer from benign nodules at early stages with sufficient confidence, therefore false positives undergoing surgical resection frequently occur. A molecular approach to the diagnosis of lung cancer through the analysis of exhaled breath could greatly improve the specificity of imaging procedures. A biomarker-driven approach to signs or symptoms possibly due to lung cancer would represent a complementary tool aimed at ruling out (with known error probability) rather than diagnosing lung cancer. Volatile and nonvolatile components of the breath are being studied as biomarkers of lung cancer. Breath testing is noninvasive and potentially inexpensive. There is promise that an accurate lung cancer breath biomarker, capable of being applied clinically, will be developed in the near future. In this article, we summarize some of the rationale for breath biomarker development, review the published literature in this field and provide thoughts regarding future directions.