The interpolation fitting method using cubic splines to represent a Ge resistance thermometer characteristic is compared to the commonly used least‐squares fitting method. For a ``poorly behaved'' thermometer characteristic, the spline function is demonstrated to give considerably better interpolation accuracy. For a ``well‐behaved'' thermometer characteristic, the spline function gives comparable or slightly better accuracy. Also, it is shown that by choosing a more efficient interval spacing distribution for the calibration points, one may obtain this improved accuracy even with fewer calibration points.