The International Practical Temperature Scale has a lower temperature limit of 90.18 K; measurements below this temperature are commonly made in terms of the various "national" scales. All of these scales differ significantly from the thermodynamic scale and exhibit undesirable variations in the size of the degree as a consequence; the national scales, moreover, are comparison scales, which are widely considered to be less satisfactory than scales defined in terms of fixed points and interpolating instruments. Various proposals for replacing this mixture by a single scale have been or are being developed; these are all based, in the 12 K to 90 K range, on the results of a comparison of four national scales recently carried out at Moscow and Teddington.This paper attempts to analyze the reproducibility of some of these proposed scales and their relation to the current best estimate of the thermodynamic scale. The significant lack of reproducibility found for these proposed scales below 50 K may either be tolerated, or alternatively must be reduced by further restrictions on thermometer characteristics. These proposed scales exhibit systematic variations in the size of the degree with temperature. These variations can be substantially reduced by modification of the mathematical treatment of the experimental data on which the scales are based; one such modification is presented here.Further analysis of the behavour of current and proposed scales will be enormously facilitated if the results of additional, highly accurate comparisons over wide temperature ranges and involving many types of thermometers become available.* The CCT64 table ([4] and section I11 below) is an analytical function of resistance ratio against temperature (for a hypothetical platinum resistance thermometer) based on five determinations, from four laboratories, of the thermodynamic scale. The resistance ratio W of a thermometer is its resistance at temperature T divided by its resistance a t the ice point (273.15 I().