A stronger argument is perhaps to be found in inscriptions which contain the phrase _legio … et auxilia eius_.[146] It could be wished that these texts were more numerous and more precise, but they support the supposition that some connexion existed between each legion and a definite group of auxiliary regiments better than any evidence previously adduced. The connexion, however, must have been very slight and easily broken. Dr. Hardy has pointed out that although three out of the four legions stationed in Germania Superior in 70 left the province for good during the following thirty-five years, there is abundant evidence that nothing like the same proportion of the auxilia stationed in the province accompanied them.[147] It is also clear that, in the second century at any rate, the number of auxilia attached to any legion was not fixed in accordance with any general principle, but depended upon the exigencies of the local situation on each frontier. A reference to the list of provincial garrisons contained in the appendix will show that whereas there are not likely to have been more than three thousand auxilia apiece to each of the three legions of Pannonia Superior, there were probably thirty thousand to be divided among the three legions of Britain, while in Dacia there was only one legion with something approaching twenty-five thousand auxilia. Still, with these reservations, it seems possible enough that the auxilia were always considered as in some sense dependent on the legions, and that where several legions were stationed in the same province, an arrangement was made dividing the auxilia into a corresponding number of groups, each of which was for certain purposes attached to a particular legion.[148]