Given a linear time-invariant controllable and observable multivariable system ẋ=Aˆ+Bˆu, y=Cˆswithminputs andpoutputs, it is possible to assign max (m,p) poles of the system (Davison 1970, Davison and Chatterjee 1971) arbitrarily close to desired locations using constant gain output feedback. This paper provides a new proof of Davison's theorem on pole-placement and leads to a design procedure which has certain advantages over Davison's method. In some cases, more than max (m, p) poles can be assigned arbitrarily. Assuming the system is output stabilizable, a least square design technique is outlined to approximate the desired pole locations when it is not possible to place all the poles.