My work on multiple degrees of freedom (DOF) input started during my graduate years at the University of Toronto. The best document I have written so far on the topic is my Ph.D. Thesis, in which I systematically examined human performance as a function of design variations of a 6 DOF control device, such as control resistance (isometric, elastic, and isotonic), transfer function (position vs. rate control), muscle groups used, and display format (silk cursor). Some of these studies are summarized in a paper published in Computer Graphics.
Given multiple (6) degrees of freedom, how well can a user operate all of them? Does it depend on the type of device? To this end, I have investigated people's ability to coordinate multiple degrees of freedom, based on three ways of quantification: simultaneous time-on-target (HFES97) , error correlation (HFES97), and efficiency (CHI'98).
Closely related to multi-degree of freedom input are manipulation and navigation in 3D interfaces. There is a different requirement between manipulation and navigation in terms integration and separation of multiple degrees of freedom. For manipulation research, see my Ph.D. Thesis. For 3D navigation research, see our paper In Search of the "Magic Carpet". With the common 3D techniques (e.g. stereoscopic display, occlusion, perspective projection, etc.), how well can we perform in the depth dimension in comparison to the horizontal and the vertical dimension? My IEEE Trans. SMC (July 1997) paper, Anisotropic human performance in six DOF tracking, offers one answer to such a question. Interestingly, there is also a difference between horizontal and vertical performance.