Spectral analysis of the biweekly measurements of line lengths from the 12-baseline network at Pearblossom, California, reveals that the precision of the prototype two-color Geodimeter is &sgr;=<a2+D2b2>1/2, where a=0.3 mm and b=0.12 ppm and D is the length of the baseline. In contrast, the computed precision is overestimated by 15--20% when a secular trend is fit to 15 months of measurements. By using spectral analysis, the variance &sgr;2 can be separated into two components: the frequency independent instrumental precision and a frequency dependent part. Evaluation of the same data with regard to their dependence upon fluctuations in weather reveal that at most, weather makes a second-order contribution of 0.01 ppm to the observed strain changes. This contribution is a factor of 5--10 times smaller than the instrumental precision. Thus the observed strain changes at Pearblossom reported by Langbein et al. (1982) during 1981 are not due to systematic errors resulting from an improper model of refraction in the atmosphere. ¿ American Geophysical Union 1987