Aren't they proportional by a square? If you have a 0-5V PWM signal at some frequency, the duty cycle is essentially the percentage of that p-p voltage (e.g., 50% duty cycle gives an effective (I think this is the word. The voltage you see for all practical purposes) voltage of 2.5V, 75% duty cycle gives 3.75V effective voltage) (PWM duty cycle is essentially the percentage of the period the signal is high or on). If I'm not mistaken, torque is linearly proportional to the power you deliver it. Power is V*I or V^2/R, and your coil resistance ideally is constant while powered, so you'll get a power proportional to the square of any given voltage you apply to the coil. So, I would say torque is directly proportional to the square of your voltage seen by any given PWM duty cycle (with some constant multiplier having to do with your motor, driver, signal levels, etc).
I'm not sure though. Just throwing out some ideas.