In order to explore the issue of accuracy variability of fusing probe vehicle data and fixed detectordata，the difference and feature，as well as the mechanisms，of link travel speeds which estimated from both dataare discussed，two kinds of dynamic travel speed data in Nagoya，Japan，that observed from probe vehicles andfixed point observed VICS ( Vehicle Information Communication System) data，are compared and explored． Theconsistency and differences of link travel speed estimated from two data resources are analyzed using regressionmodel． The influence factors are further explored． Ｒesults reveal that great differences in traffic information do existbetween two data sources，due to the main reasons such like the sample size of probe vehicle and the uncertainty insampled probe vehicles，the delay transmission of VICS，and the manual setting upper bound travel speeds of VICSsystem． Noted that the error mechanisms for different sections are distinguished with each other，it is easy to under-stand that using a general fusion algorithm cannot always improve fusing accuracy for all road links if ignoring errorspropagation during data collection and processing．