OBJECTIVE Absorbed dose beam quality conversion factors (k(Q) factors) were obtained for 27 different types of ionization chamber. The aim was to obtain objective evidence on the performance of a wide range of chambers currently available, and potentially used for reference dosimetry, and to investigate the accuracy of the k(Q) calculation algorithm used in the TG-51 protocol. METHODS Measurements were made using the 60Co irradiator and Elekta Precise linac facilities at the National Research Council of Canada. The objective was to characterize the chambers over the range of energies applicable to TG-51 and determine whether each chamber met the requirements of a reference-class instrument. Chamber settling, leakage current, ion recombination and polarity, and waterproofing sleeve effects were investigated, and absorbed dose calibration coefficients were obtained for 60Co and 6, 10, and 25 MV photon beams. Only thimble-type chambers were considered in this investigation and were classified into three groups: (i) Reference chambers ("standard" 0.6 cm3 Farmer-type chambers and their derivatives traditionally used for beam output calibration); (ii) scanning chambers (typically 0.1 cm3 volume chambers used for beam commissioning with 3-D scanning phantoms); and (iii) microchambers (very small volume ion chambers (< or = 0.01 cm3) used for small field dosimetry). RESULTS As might be expected, 0.6 cm3 thimble chambers showed the most predictable performance and experimental k(Q) factors were obtained with a relative uncertainty of 0.1%. The performance of scanning and microchambers was somewhat variable. Some chambers showed very good behavior but others showed anomalous polarity and recombination corrections that are not fully explained at present. For the well-behaved chambers, agreement between measured and calculated k(Q) factors was within 0.4%; for some chambers, differences of more than 1% were seen that may be related to the recombination/polarity results. Use of such chambers could result in significant errors in the determination of reference dose in the clinic. CONCLUSIONS Based on the experimental evidence obtained here, specification for a reference-class ionization chamber could be developed that would minimize the error in using a dosimetry protocol with calculated beam quality conversion factors. The experimental k(Q) data obtained here for a wide range of thimble chambers can be used when choosing suitable detectors for reference dosimetry and are intended to be used in the upcoming update/addendum to the AAPM TG-51 dosimetry protocol.