Learning in the temporal bisection task: Relative or absolute?