Haptic technology is more and more widely used to improve human interaction with devices, for example in touch screens of smartphones that vibrate when touched. Another application is haptic-tele-manipulation where a human controls a slave manipulator (e.g. a surgical-robot) by using a master device (e.g. a joystick). In haptic-tele-manipulation, forces measured at the slave manipulator are fed back to the human operator is order to improve task performance. Currently, human force perception is neglected in haptic system design, implicitly assuming perfect force perception, requiring high performance of the master and slave devices. The goal of this thesis is to identify key factors that influence isometric (static) force perception, and to develop metrics and computational models that quantify and predict this influence. To reach this goal, isometric force reproduction experiments were performed in which subjects were asked to actively generate the target and reproduction force using the same hand. In this thesis, we analyzed the influences of different factors using the same experimental protocol. Subjects were asked to perform series of two interchanging trials: matching an onscreen force in magnitude and direction (target trial), and subsequently reproducing the same force vector without visual feedback (reproduction trial). The difference in forces exerted during the target and reproduction trial is called the force reproduction error. In chapter two, we analyzed the effect of force magnitude on the force reproduction error and position reproduction error in one degree of freedom (DOF). Subjects performed force reproduction tasks at different force levels (10N – 160N, with 30N increments), against a fixed handle, and performed a position reproduction task against a haptic manipulator, which applied constant opposing forces. Subjects reproduced too high forces for low force levels (<40N) and too low forces for high force levels (>130N). No effect of force level on the position reproduction error was found. If the force reproduction error is exclusively caused by the reafference feedback (by the CNS predicted sensory feedback caused by self-generated forces), the force reproduction error should disappear when both target and reproduction force are self-generated. The results of this study show an effect of force level on the force reproduction error, indicating that reafference feedback is not the sole factor in force reproduction tasks. In chapter three, we analyzed the effect of force direction and arm-posture on the force reproduction error in the horizontal plane. The force reproduction protocol was performed in eight force directions and in four arm postures at a force level of 10N. Results showed that the force reproduction error depends on force direction. The orientation of the ellipses fitted through the reproduction forces changed with arm posture and the least accurate direction aligned with the shoulder in all postures. For each of the four arm postures, a joint torque scaling model based on arm biomechanics was fitted to the other three postures and was shown to accurately explain the reproduction ellipse. This chapter shows that force reproduction depends on force direction and arm posture, which corresponds with our model based on arm biomechanics and suggests that the force reproduction errors at the endpoint originate at the joint torque level. Chapter four assesses the effect of force magnitude on the force reproduction error in the horizontal plane. Three groups of subjects performed the force reproduction protocol in eight force directions and in two arm postures at three force levels (group1: 10N, group2: 40N and group3: 70N). Results show that the orientation of the reproduction ellipses changes with arm posture but not with force level, indicating that the arm biomechanics (i.e. arm orientation) affect the directional and force effects. Additionally, the results show that force level affects the force reproduction error differently depending on the force direction, i.e. increased force magnitude increases the errors in the direction of the shoulder and decreases the errors in the perpendicular direction. To incorporate the force level dependency of the force reproduction error, a novel joint-torque-dependent joint-torque-scaling model is developed, which allows us to accurately predict the force reproduction errors in the horizontal plane between 10N and 70N. In chapter five, we present an experimental study to examine whether the systematic errors result from an incorrect representation of forces or incorrect execution of correctly represented forces. In this experiment we asked the subjects to reproduce the same magnitude of force (5N or 15N) in either the same direction (reference trial) or in a 90?-counter-clockwise-rotated direction (CCW trial). We hypothesize that if the force reproduction errors arose through an execution error, subjects would obtain the correct representation of target force but reproduce this incorrectly in the CCW trials, resulting in an identically oriented reproduction ellipse as in the reference trials. However, if these systematic errors arose through an incorrect representation, then we hypothesize that subjects would produce a 90?-rotated ellipse relative to the reference trials. The absence of any rotation in our results demonstrates that the forces were represented correctly, and the systematic errors arise as execution errors during reproduction. Moreover, the results suggest that the sensorimotor system does not attempt to match the sensory percept or the motor commands, but instead develops an internal representation of the forces. In chapter six the main results of the thesis are discussed, where we present three overall conclusions: 1) Human force perception comprises systematic errors; 2) Systematic errors in force perception originate at joint level and are predictable; 3) The systematic errors in force perception are execution errors. Based on these overall conclusions we provide two guidelines for haptic system design (H) and two guidelines for neuroscience (N): H1) Take the accuracy of human force perception into account to make haptic devices more affordable; H2) Compensate for the systematic errors in human force perception; N1) The errors for different modalities originate in different reference frames; N2) The central nervous system does not simply compare sensory information or motor commands to control the body. The work described in this thesis provides novel insight in the accuracy of human force perception, presents a model that can accurately predict the force reproduction error, and provides the first steps in determining where the errors originate.