Many cellular processes require or depend on mechanical feedback from the cell's micro-environment. For example, the local organization of the actin cytoskeleton is known to adapt to the physical properties of the extracellular matrix. To gain better understanding of the mechanisms controlling this process, we have studied the particular case of peripheral stress fibers, thick concave contractile bundles of actin filaments. We have employed micropatterned adhesive substrates to constrain cell shape and adhesion layout. We have then used fluorescence microscopy, traction force microscopy and cantilever-based force measurements to experimentally characterize the assembly and mechanical tension of these peripheral stress fibers. We combined these experimental data with numerical simulations of cell shape and cell tension to propose a general mechanical model of stress fiber behavior. We first identified the formation of peripheral stress fibers as significant of a behavioral switch triggered by non-adhesive gaps in the substrate geometry. These bundles indeed form above a threshold distance of 3-4µm, necessary to stimulate adhesion maturation, and lead to a redirection of cellular tension parallel to the cell edge. Additional analysis of cells on rigid and soft substrates showed that the radius and tension of peripheral bundles weakly increase with spanning distance. Thus both the cytoskeleton local organization and its mechanical properties can be controlled to some extent by non-adhesive features in substrate geometry. The peripheral fibers' concave curvature results from the balance of the line and surface tensions. However, the nature of these forces, and in particular the contribution of myosin-dependent contractility, are not clear. To get insight into this force balance, we inhibited myosin activity and monitored the shape and tension responses. We found that myosin inhibition led to a decrease in the traction forces and an increase in arc radius, indicating that line tension dropped to a lesser extent than surface tension. These results suggest a myosin-independent component in the tension of peripheral arcs. We propose a simple physical model in which the line tension is the sum of a myosin-dependent active component and of a passive elastic component. Numerical simulations of this model reproduce well the measured shape dynamics and allow estimating the relative contributions of elasticity and contractility to the arc line tension. We then developed an alternative setup to directly probe the elasticity of semi-isolated peripheral fibers, using micromanipulation techniques. A soft cantilever was used to measure the force with which the fiber resisted a transverse deformation. The overall mechanical response of the fiber can be represented by a 3-parameter-solid model, i.e. a spring in series with a Kelvin-Voigt unit. The response was separated into two regimes: first a fast traction during which elastic response dominated, followed by a viscoelastic relaxation. The elastic and viscoelastic coefficients were extracted from the force-elongation curves and from the elongation velocity during relaxation. Inhibiting myosin activity decreased the active pre-tension, but had no clear effect on the elasticity. Our results show that myosin is an important actor, but not the only one, in the response of peripheral fibers to external mechanical constraints (active deformation) or geometrical constraints (limited adhesive area).