SS Total is the amount of error revealed by the empty model (the mean); it is the total area of the squared residuals based on the distance of each score from the mean.