Five young male subjects performed a simulated raking task by pushing and pulling 14 different rake handles at 60N and 110N simulated soil resistance in a random order. The raking action was performed at a self-selected uniform pace over a 60 cm stroke length with initial, middle and final phases marked. During this action the raking force and the angle of the rake was measured by a load cell and a potentiometer respectively. The posture was recorded with a video cassette recorder. The posture and force values were used for the determination of the spinal compression load. In rake pulling the 13 modified rake handles generated a spinal compression of only 20% to 50% of the straight handle, whereas in rake pushing the modified handles generated compression up to five times that of the straight handle. The compression generated with the straight handle never reached the action limit, whereas those of the 13 modified handles rarely stayed within the action limit. Therefore, a straight handle is considered the handle of choice.