At many sites in Ireland where the Normans are known to have built castles at an early period of the invasion there are no mottes to be seen now. It is probable that where the Norman conquerors had both money and time at their disposal they built stone keeps from the first, and that the motte-castles, with their wooden towers or _bretasches_, were built in the times of stress, or were the residences of the less wealthy under-tenants. But we know from documents that even in John's reign the important royal castle of Roscrea was built with a motte and bretasche,[1048] which proves that this type of castle was still so much esteemed that we may feel reasonably certain that when Giraldus speaks of "slender defences of turf and stakes" he does not mean motte-castles, but mere embankments and palisades.[1049] But there is another reason for the absence of mottes from some of the early Norman castle sites. Those who have examined the castles of Wales know that it is rare to find a motte in a castle which has undergone the complete metamorphoses of the Edwardian[1050] period. These new castles had no keeps, and necessitated an entire change of plan, which led either to the destruction of the motte or the building of an entirely new castle on a different site. The removal of a motte is only a question of spade labour, and many sites in England can be pointed out where mottes are known to have existed formerly, but where now not a vestige is left.[1051] There are many other cases where the Edwardian castle shows not a trace of any former earthworks, but where a motte and bailey a little distance off probably represents the original wooden castle.[1052]