Templating of preoperative radiographs is routinely recommended prior to knee arthroplasty. We performed this study to assess the reproducibility and accuracy of the templates for three commonly used knee implants (PFC, Kinemax, Scorpio). Six lower limb surgeons templated 10 patients for each of the three designs. The inter- and intra-observer reliability and accuracy was calculated. There was marked variation in the reliability of the templating with the tibial insert scoring better than the femoral and the Kinemax being the most reproducible of the three. In general, the intra-observer scores (kappa=0.57-0.81) were better than the inter-observer ones (kappa=0.21-0.60). The Scorpio was the most accurately templated of the three implants, with the percentage correlating with what was actually implanted ranging from 55% to 62% for the femur and 72% to 75% for the tibia, with no templated sizes more than one size different from the actual implant. The other implants ranged from 38% to 42% for the femur and 53% to 58% for the tibia with both having up to 3% more than 1 size difference from the actual implant. We believe that the use of templating in total knee arthroplasty should be interpreted with caution and we urge the development of more accurate prosthesis sizing techniques.