Cell migration results from forces generated by assembly, contraction, and adhesion of the cytoskeleton. To address how these forces integrate in space and time, novel assays are required that allow spatial separation of the different force categories. We used micro-contact printing of fibronectin on glass substrates to study the effect of adhesion patterns on fish epidermal keratocytes locomotion. Cells migrated at similar speeds on homogeneously adhesive substrates and on patterns with 5 microm-wide adhesive stripes interleaved by non-adhesive stripes with a width varied between 5 and 13 microm. The leading edge protruded on adhesive stripes and lagged behind on non-adhesive stripes. On patterns with non-adhesive stripes wider than 13 microm cells halted, although the lamellipodium did not collapse. High correlation was found between the widths of protruding and lagging edge segments and the widths of the underlying stripes. We explain our data by the force balances between actin polymerization, contraction and adhesion on fibronectin stripes; and between actin polymerization, contraction and lamellipodium-internal elastic tension on non-adhesive stripes. We tested our model further by blocking lamellipodium actin network contraction and polymerization. In both experiments we observed that cells eventually lost their ability to move. However, the two perturbations induced distinct morphological responses. The data suggested that forces powering forward motion of keratocytes are largely associated with network assembly whereas contraction maintains cell polarity. This study establishes spatially selective adhesion substrates and cell morphological readouts as a means to elucidate the mechanical balance between substrate adhesion and cytoskeleton-internal tension in cell migration.