With modern space geodetic techniques the motions of points on the surface of the solid Earth can be determined at accuracy of centimeters or even millimeters per year. Precision of results from data analysis is always very high, but results of various individual analyses may not be consistent with each other very well and the differences between results may not always be accounted for by analysis formal errors. For example, as the motion of the Shanghai collocation site relative to the Eurasia plate is concerned, in Shanghai Observatory, it is in the northeast according to the GPS group, but it is in the southeast according to the VLBI group as well as the SLR group. Is this technique dependent? software dependent? or just analysis dependent? These are complicated questions to answer, but are really tantalizing.

After individual solutions of the terrestrial reference frame from various techniques are transformed to the ITRF96, the motions of the Shanghai collocation site relative to the Eurasia plate are compared. The horizontal rates are around 10 mm/yr. The azimuths of these rates are from 20 to 120 degree for GPS solutions, from 60 to 100 degree for VLBI solutions, and from 90 to 120 degree for SLR solutions. So the motions from various techniques are roughly east-ward, but the scatter is obvious especially in the azimuth. It should be noticed that the differences among individual solutions of an individual technique are obvious too, which are even more obvious for GPS than those for various techniques.

The motions of several decades collocation sites are also compared in the same way. For most sites the motions from individual solutions are generally clustered together and the scatter in azimuth is more outstanding than in horizontal rate. Considering that the magnitude of the east-ward as well as the north-ward component motion is very small, which is usually several millimeters per year, so it is very sensitive to error sources. And that the horizontal rate is the square-root of the square-sum of components, while the azimuth of the rate is determined from the arc-tangent function of the ratio of east to north component, 10% variation in one component will result in more outstanding change in the azimuth than in the rate. Therefore the indication of our comparison of motions of collocation sites are at least partly understandable.

Both hardware and software of modern space geodetic techniques are the representation of the state of art. These along with the observations are by most part shared in the whole scientific society of the world. Whenever the differences among results of various techniques are concerned, it is more likely to be treated as analysis dependent rather than techniques dependent.