Skeletal muscle is the engine that produces force to power movement in humans and animals alike. To date the invasive nature of obtaining muscle-tendon forces in humans has limited our understanding of muscle function during coordinated locomotor tasks. Phenomenological, Hill-type models of skeletal muscle are often used, providing estimates of a muscle’s force as a function of its activation state, force-length, and force-velocity properties. However, few studies have examined the accuracy of whole muscle-tendon forces obtained from such models during in vivo motor tasks. The goal of my thesis was to develop, test, and refine methods to better quantify muscle mechanical output in humans, using ultrasound and electromyographic recordings, together with advanced Hilltype models. My first study developed techniques to non-invasively estimate in vivo Achilles tendon forces. I used ultrasound-based measures of tendon length and tendon mechanical properties to determine forces during cycling. In my second study, I compared gastrocnemii forces, predicted from a traditional Hill-type model with one contractile element, to force estimates derived from ultrasound-based tendon length changes. Because the traditional Hill-type model fails to account for variable activation states of different fibre types, I additionally tested a two-element model that includes both slow and fast contractile elements. I found that Hill-type models predicted 31-85% of the cyclists’ gastrocnemii forces across a range of conditions elicited, producing results comparable to those reported in animal models. Further, at higher cadences, the two-element model better estimated forces because it accounted for the increased recruitment of fast fibres. Traditional Hill-type models also neglect dynamic shape changes in contracting muscles, which may be important in modulating the velocities at which fascicles operate. My third study compared predictions of muscle architecture (fascicle lengths and pennation angles) generated from a 1D Hill-type model and additionally from 2D and 3D geometric models that allowed dynamic shape changes to occur. I found that the 1D model provided predictions of muscle architecture that were similar to the predictions of 2D and 3D models and that muscle shape changes and fascicle velocities were more closely linked to force than activation. Taken together, this research provides a non-invasive approach for studying in vivo muscle-tendon mechanics and testing the predictions of Hill-type models.