In positron emission tomography (PET), the large number of lines of responses in three-dimensional (3D) acquisition mode creates a high volume of sinogram data and increases reconstruction time in iterative reconstruction. We tried to decrease sinogram data volume by reducing the number of views using angular compression and then evaluated the accuracy of this mashed mode. Three methods were compared, conventional mode (CONV), X2 mashed mode (X2: two adjacent projection angles are added together), and X4 mashed mode (X4: four adjacent angles added). A point source of (18)F was used to measure spatial resolution. A hot spot phantom made of 6 hot spheres (10-38 mm in diameter) within water of 20 cm in diameter was scanned to evaluate the recovery coefficient (RC). A lung-heart-liver phantom made of homogeneous radioactive myocardium, a spherical hot mass in the lung (10 mm in diameter), and background activity in the liver was scanned to evaluate the homogeneity of the myocardial wall. The quality of the reconstructed images was evaluated in terms of the normalized mean square error (NMSE), Bull's eye map, profile curve, and peak value of the spherical hot mass. The reconstruction times of X2 and X4 were one-half and one-quarter, respectively, of that of CONV. In terms of spatial resolution, FWHM of CONV, X2, and X4 were, 4.26, 4.33, and 4.48 (mm) at the center, 4.81, 5.68, and 8.73 tangentially, and 8.01, 8.19, and 8.27 radially at R=200 mm, respectively. RC was similar for all methods. The NMSE values of X2 and X4 compared with CONV were 0.0003 and 0.0014, respectively. In the hot mass, these methods showed almost the same profile curves, although the peak value of X4 was only -1.95% less than that of CONV. Although the result of spatial resolution of X4 was slightly degraded, image quality and physical performance were good. Therefore, the X4 mashed mode used with angular compression was considered clinically useful.