Model 6711 I-125 seeds are commonly used in permanent brachytherapy implants. While recommended dose distribution parameters for these sources have been published by AAPM TG-43, several investigators have recently questioned the presence of seemingly unphysical fluctuations in the anisotropy function data. Seeking to understand these, we measured the dimensions of eight model 6711 seeds radiographically, and made a new experimental determination of the anisotropy function and factor using thermoluminescent dosimeters (TL-100 chips and minicubes) in a solid water phantom. It was found that variations in seed capsule end weld thickness, and movement within the capsule of the Ag rod onto which the 125I is adsorbed, were present for all sources and thus contributed to experimental uncertainty. Averaging results from two different sources, from data acquired at the same time from diametrically opposed quadrants of the phantom, and from repeated measurements yielded anisotropy function values similar to those of TG-43 but characterized by greater precision (less than or equal to 3% for radial distance r less than or equal to 3 cm and less than or equal to 1% for r greater than or equal to 4 cm), an absence of sharp fluctuations, and reduced magnitudes at r = 1 cm. The measured values can be well represented by an analytic function similar to that proposed by Furhang and Anderson to fit the TG-43 data. Values of the anisotropy factor derived from this function by integration exhibit little variation with r, in agreement with earlier diode data but in contrast to TG-43 data, and can also be represented by an analytic function. Finally, a difference in TLD chip and minicube reproducibility, observed here and by earlier investigators, is explained by reference to recent work (done concurrently with ours) as stemming from variations in dosimeter orientation and automated reader positioning/heating for minicubes. (C) 2000 American Association of Physicists in Medicine. [S0094-2405(00)02608-0].