In this paper we are concerned with the problem of sensory motor coordination of a robotic finger in order to evoke rolling maneuvers in a force-positioning task on a flat rigid surface. We use two different approaches to modeling the reaction of the soft fingertip with the contacted surface. Firstly, we assume that the environment imposes a purely kinematic rolling constraint on the end-effector motion in the tangent direction of the contacted surface. This implies no energy transfer or dissipation between the fingertip and the environment due to frictional forces. On the other hand, we assume that it is feasible for the fingertip to slip in which case pure rolling motion could be disturbed. The two different models are subsequently used to show by simulation that control laws, which have been designed on a rolling constraint dynamic model for frictional forces, fail to perform rolling in various environments. An extra control input that uses a reference rolling trajectory that is state dependent is proposed, which, if superimposed on a conventional force-position control law, can achieve rolling even on a surface with low friction characteristics. The proposed feedback signal does not utilize the modeling information in the control formulation, and thus permits easy implementation. Finally, the total controller is shown to achieve asymptotic convergence to the desired force-positioning task by simultaneously evoking pure rolling motion for the fingertip.