Although the California geodimeter network has been surveyed repeatedly over a period of almost 15 years, three different procedures have been employed successively in the periods 1959–1964, 1964–1969, and 1969–1974. Direct comparison of the last two procedures suggests a discrepancy of about 13 mm + 0.8 ppm in line length measurement, the post-1969 measurements being longer. An analysis of the temperature measurements used to compute the refractivity correction suggests a discrepancy between the line length measurements of 1959–1964 and 1964–1969 amounting to perhaps 0.8 ppm, the 1964–1969 measurements being longer. It would appear that the discrepancy between the 1959–1964 and 1969–1974 procedures might lead to an apparent shortening of geodimeter lines by 13 mm + 1.6 ppm, which for an average length line (23 km) amounts to almost 50 mm. The bias is statistical in nature and cannot be removed from individual measurements. The observed discrepancies are plausible consequences of the usual nocturnal ground level temperature inversion, but definite proof of a causal relation is lacking. The bias is capable of accounting for several anomalies in the interpretation of the geodimeter data (e.g., the apparent dilatancy near Parkfield reported by Cherry and Savage).