Five high-sensitivity metal oxide semiconductor field effect transistor dosimeters in the TN-502 and 1002 series (Thomson Nielsen Electronics Ltd., 25B, Northside Road, Ottawa, ON K2H8S1, Canada) were evaluated for use in the mammography x-ray energy range (22-50 kVp) as a tool to assist in the documentation of patient specific average glandular dose. The dosimeters were interfaced with the Patient Dose Verification System, model No. TN-RD 15, which consisted of a dosimeter reader and up to four dual bias power supplies. Two different dual bias power supplies were evaluated in this study, model No. TN-RD 22 in high-sensitivity mode and a very-high sensitivity prototype. Each bias supply accommodates up to five dosimeters for 20 dosimeters per system. Sensitivity of detectors, defined as the mV/C kg(-1), was measured free in air with the bubble side of the dosimeter facing the x-ray field with a constant exposure. All dosimeter models' angular response showed a marked decrease in response when oriented between 120 degrees and 150 degrees and between at 190 degrees and 220 degrees relative to the incident beam. Sensitivity was evaluated for Mo/Mo, Mo/Rh, and Rh/Rh target-filter combinations. The individual dosimeter model sensitiVity was 4.45 x 10(4) mV/C kg(-1) (11.47 mV R(-1)) for TN-502RDS(micro); 5.93 x 10(4) mV per C kg(-1) (15.31 mV R(-1)) for TN-1002RD; 6.06 x 10(4) mV/C kg(-1) (15.63 mV R(-1)) for TN-1002RDI; 9.49 x 10(4) mV per C kg(-1) (24.49 mV R(-1)) for TN-1002RDM (micro); and 11.20 x 10(4) mV/C kg(-1) (28.82 mV R(-1)) for TN-1002RDS (micro). The energy response is presented and is observed to vary with dosimeter model, generally increasing with tube potential through the mammography energy range. An intercomparison of the high-sensitivity mode of TN-RD-22 was made to the very-high sensitivity bias power supply using a Mo/Mo target-filter. The very-high sensitivity-bias power supply increased dosimeter response by 1.45 +/- 0.04 for dosimeter models TN-1002RD and TN-1002RDM. The responses of all dosimeter models were found to be linear for tube potentials of between 24 and 48 kVp. Dosimeters showed a reproducibility varying from 15.5% to 31.8%. depending on the model of dosimeter. Micro MOSFETS model Nos. TN-1002RDS and TN-1002RDM used in conjunction with their respective high-sensitivity and ultrahigh-sensitivity bias supplies provided the highest sensitivity response of the models evaluated. Either micro MOSFETS model No. TN-1002RDS or TN-1002RDM used in conjunction with the appropriate bias supply provide the best choice for clinical mammography applications. Under these conditions, MOSFET dosimeters can provide a viable option as a dosimeter in the mammography energy range (22-50 kVp). The clinical application of MOSFET dosimeters must take into account the energy dependence and reproducibility to ensure accurate measurements.