Sea-ice movement is mainly driven by stress from surface wind and ocean currents. The spatial inhomogeneity of these forces causes internal sea-ice stress gradients, which eventually cause ice convergence or divergence, i.e. to compress or break-up. This sea-ice deformation occurs across a broad range of spatial scales. Most noticeable, however, are linear kinematic features (LKFs) that have lengths of hundreds to thousands of kilometers and a typical lifetime of several days. Sea-ice deformation is an important process for: (1) sea-ice mass balance due to new ice production; (2) brine rejection into the ocean; (3) regulation of ocean-to-air heat flux; and (4) altering the air and water drag coefficients. Coupled sea-ice–ocean models can reproduce some aspects of sea-ice drift. Detailed comparisons between satellite remote-sensing data with model results, however, reveal big differences in the shape, frequency of occurrence and spatial distribution of LKF-like features. Their representation strongly depends on the physics of the sea-ice model. Here we study the Arctic simulations carried out with the MIT general circulation model using a viscous–plastic sea-ice rheology with an elliptical yield curve. Similar sea-ice models are broadly used also for fully coupled climate simulations. The simulated sea-ice motion and deformation are compared with estimates from synthetic aperture radar satellite retrievals from the RADARSAT Geophysical Processor System (RGPS). This study addresses the following questions: (1) How does the spatial distribution and shape of LKFs depend on model resolution? (2) Can the model parameterization be improved to better represent the observed large-scale sea-ice deformation field? (3) How well is the seasonal cycle of sea-ice deformation represented in the model? (4) Is the observed relationship between sea-ice thickness and deformation adequately represented in the model? Even though the viscous–plastic dynamic sea-ice model is able to produce what appears to be LKFs, the orientation and spatial density differ considerably from the observations. In addition, the LKFs occur less frequently in the simulations, especially in the seasonal sea-ice zone. This result holds for integrations with both 9 km and 18 km horizontal grid spacing, even though the 9 km integration has more and better-defined LKFs. By changing the sea-ice stress parameterization, the large-scale ice-deformation distribution can be made more similar to that of the RGPS data but differences for smaller-scale deformation features still persist.