Flying squirrels are well known for their ability to glide between trees at the top of a forest canopy. We present experimental performance and behavioural evidence that flight in flying squirrels may have evolved out of a need to control landing forces. Northern flying squirrels were filmed jumping from a horizontal branch to a much larger vertical pole. These were both slightly compliant (less than 1.9 mm N(-1)), and instrumented using strain gauges so that forces could be measured. Take-off and landing forces were both positively correlated with horizontal range between 0.5 and 2.5 m (r=0.355 and r=0.811, respectively, P<0.05), but not significantly different to each other at each range tested. Take-off forces ranged from 1 to 10 bodyweights, and landing forces were between 3 and 10 bodyweights. Glide angles increased rapidly with horizontal range, approaching 45 degrees at 3 m, above which they gradually decreased, suggesting that northern flying squirrels are optimised for long distance travel. We show that northern flying squirrels initiate full gliding posture at ranges of less than 1 m, without landing any higher than an equivalent ballistic projectile. However, this gliding posture enables them to pitch upwards, potentially stalling the wing, and spreads the landing reaction force over all four extended limbs. At steeper approach angles of close to 45 degrees , flying squirrels were unable to pitch up sufficiently and landed forelimbs first, consequently sustaining higher impact forces. We investigate four hypotheses to explain the origin of flight in these animals and conclude that the need to reduce landing impact forces was most likely to have stimulated the development of aerial control in flying squirrels.