We use a short-baseline network of braced monuments to investigate site-specific GPS effects. The network has baseline lengths of ∼10, 100, and 1000 m. Baseline time series have root mean square (RMS) residuals, about a model for the seasonal cycle, of 0.05–0.24 mm for the horizontal components and 0.20–0.72 mm for the radial. Seasonal cycles occur, with amplitudes of 0.04–0.60 mm, even for the horizontal components and even for the shortest baselines. For many time series these lag seasonal cycles in local temperature measurements by 23–43 days. This could suggest that they are related to bedrock thermal expansion. Both shorter-period signals and seasonal cycles for shorter baselines to REP2, the one short-braced monument in our network, are correlated with temperature, with no lag time. Differences between REP2 and the other stations, which are deep-braced, should reflect processes occurring in the upper few meters of the ground. These correlations may be related to thermal expansion of these upper ground layers, and/or thermal expansion of the monuments themselves. Even over these short distances we see a systematic increase in RMS values with increasing baseline length. This, and the low RMS levels, suggests that site-specific effects are unlikely to be the limiting factor in the use of similar GPS sites for geophysical investigations.