Human-computer interaction (HCI) has been traditionally analyzed in terms of two-way information exchange. I review empirical evidence suggesting that, if approached from this perspective, mobile devices are inherently inferior in comparison to desktop computers. I then make a case for embodied interaction--that is, leveraging users' ability to perceive and represent the structure of the proximate environment and their capacity to transform tasks by means of action. While there is a lot of potential in this interaction paradigm, an outstanding scientific problem is how to measure information capacity in a case where performance is not stimulus-bound; that is, when it is not dictated by the external environment. I present an extension of Fitts' law that covers unconstrained whole-body movement. The results of this work will provide a universal basis for comparison of novel interaction techniques. I am presently applying these ideas in the area of mobile mixed reality interfaces, aiming toward a vision where mobile users can literally sense and act through digital information.