Cognitive Biomechanical Decisions to Negotiate Unstable Branches in Fox Squirrels. RUOPP, R.*; WANG, L; LEE. S.; FULL, R.J. Univ. of California, Berkeley. rruopp@berkeley.edu Arboreal agility requires more than skilled biomechanics. Critical cognitive behavioral decisions often must be made instantly including knowledge of biomechanical capability. We challenged free-ranging Fox squirrels to jump across three elevated branch-like rods in the same plane, but perpendicular to their forward path in return for a food reward. We varied the spacing of the rods (25, 37, 50, and 62cm) and the stability of the center rod from fixed (Non-Rotating, NR) to rotating (R). We recorded behavioral decisions and leaping kinematics with four high-speed video cameras for 652 leaping trials from twelve, identifiable individuals. Using a Markov-like chain analysis, we found that as jump distance increased failures to negotiate the R rod increased. Squirrels showed exploratory testing behavior by jumping to the center R rod and then back to the starting NR rod. Variation of the strategies selected increased most after a failure. Remarkably, as gap distance increased (50 cm) and stable landing on the R rod became more challenging, squirrels showed innovation by unexpectedly leaping high onto the fixed side structures securing the rods, bypassing the R rod. Eventually squirrels learned to leap stably from the R rod to cross the entire set-up, especially at the greatest gap distances where other biomechanical options appeared more limited. With further analysis and by inclusion of additional obstacles and controls, we hope to develop a model of embodied learning and control that will serve as biological inspiration for the most agile robot yet built.