Thoracoscopy for wedge resection of lung metastases is rapidly increasing in frequency. This technique precludes bimanual palpation of the lung to locate additional lesions not seen on the surface. Finger palpation is inadequate. Implications regarding the failure to identify all metastases and the negative impact on long-term survival led us to review retrospectively the correlation between pathologic findings and imaging reports. One hundred forty-four patients who had resection of lung metastases from colorectal cancer were studied. All had chest roentgenograms and 72 had computed tomographic scans as well. Chest roentgenogram and computed tomographic reports differed in the number of nodules reported in 17 of 72 patients (24%). In 3 of 17 patients chest roentgenogram showed more nodules than computed tomography. Chest roentgenogram differed from pathologic findings at surgery in 57 of 144 patients (39%). Twenty-six of 57 patients (46%) had more lesions than chest roentgenogram detected and 31 had fewer. Computed tomographic scans differed from pathologic findings in 30 of 72 patients (42%). If one or two lesions were imaged, 12 patients had fewer cancers (some lesions were benign) and 18 had more cancers than computed tomography reported; computed tomographic scans erred 28% of the time. The inability to adequately palpate the entire lung using the thoracoscope alone markedly impairs the surgeon's ability to know if a resection of all lesions has been done. The validity of using thoracoscopy resection in the management of metastatic disease is seriously questioned other than for diagnosis.