In 2D-mode positron emission tomography (PET), scatter is either compensated by approximate methods based on the existing emission data or ignored altogether as the magnitude of the scatter fraction (SF) is on the order of 10%-20%. In clinical PET studies, however, attenuation and sophisticated scatter correction methods are required along with CT or radionuclide transmission scans. With the growing interest in small animal imaging, these correction methods are being translated to small animal scanners, but there is little scientific information about the requirements associated with smaller size objects and scatter geometries. In this study, we focused on the magnitude of the scatter through a series of scatter fraction simulations. To determine the scatter as a function of object size, we performed Monte Carlo simulations using GATE (Geant4 application for emission tomography). Models of the ECAT HR+ PET scanner (included in the GATE package) and the Siemens Inveon small animal scanner (generated by the first author) were used. Simulations were performed for several digital phantoms including the NEMA, XCAT and MOBY phantoms over a wide range of sizes. Small animal NEMA-like phantoms indicated that for cylindrical objects less than 5 cm diameters (encompassing small rats and all mice), the scatter fraction was lower than 17.5% for the 350-650 keV and 20% for the 250-750 keV windows. Similar values were obtained for the MOBY phantoms for the respective sizes. On the other hand, the scatter fraction was more than 35% for even the smallest size human NEMA-like and XCAT phantoms. These results suggest that sophisticated scatter correction methods may not be required for the indicated sizes of mice and rats.