U-value measurements made with the MoWiTT field test facility and at a commercial laboratory for four com.- mercial windows are compared with calculations made with the WINDOW program. Good agreement is found for three of the windows; for the fourth--a double-glazed window with a highly conductive frame--agreement is good between the calculations and the MoWiTT measurements, but agreement with the test laboratory is only marginal Measurements of overall diurnal performance are presented, and it is shown that, even for a north.facing window, the uncertainties in thermal performance due to solar gain effects overshadow the effects of improved U-value. The author argues the need for better methods of accounting for solar gain effects in window performance comparisons, so that the net benefits of Uvalue improvements may be correctly assessed.