One difficulty in radiation therapy is ensuring the correct placement of the radiation field so that the radiation is delivered to the diseased regions while healthy tissues are spared. Currently, field placement is assessed by producing a transmission radiograph (portal image) using the high-energy treatment beam. Often the quality of these images is poor. One factor which influences image quality is the size of the x-ray source found in medical linear accelerators. These sources can be large and thus reduce the spatial resolution of the portal images, thereby reducing our ability to detect bony anatomical landmarks. The high energy of the x rays generated in these accelerators make it impossible to measure the x-ray source using standard techniques (e.g., pin-hole camera or star-patterns). We have developed a reconstruction technique which allows quantitative measurement of these sources. Using this technique, we have investigated and compared the size and shape of x-ray sources on a total of seven accelerators. These include: (1) two Varian Clinac 2100cs, (2) two Atomic Energy of Canada Ltd. (AECL) Therac 6s, (3) a Varian Clinac 600c, (4) an AECL Therac 25, and (5) a Therac 20. The comparisons also include monitoring the size and shape of the sources over a two year period. For each of these measured sources the source MTF has been calculated at typical magnifications. It has been found that (1) the size and shape of the source spot varies greatly between accelerators of different design; (2) for accelerators of the same design, however, the source spots were similar; and (3) the spatial resolution of currently available on-line verification imaging systems is only marginally reduced by the size and shape these source spots.